0

STATISTICAL

METHODS IN CIVIL

ENGINEERING
LECTURE NOTES

PROF. DR. ÖMER YÜKSEK

KTÜ CIVIL ENG. DEPT.

TRABZON
 1

CHAPTER 1

INTRODUCTION

1.1. STATISTICS IN ENGINEERING

1.1.1. General Introduction

Probabilistic methods are with increasing frequency used in the design of civil structures such as channels,
storm surge barriers, bridges, buildings, etc. The methods are being applied directly, or are translated to
relatively simple design rules with safety coefficients. In both cases the foundation of the calculations is
given by the statistical distribution functions of the strength and load variables.
In the application of probabilistic methods, the availability of useful calculation models and adequate
statistical distributions are required. The development of calculation models has had a lot of attention during
the last years. Methods for finding the best estimate, as well as a quantification of the uncertainty in the
estimate, should be described. How to deal with uncertainties is an essential part of statistics science.
Probability and statistics are concerned with events which occur by chance. Examples include occurrence of
accidents, errors of measurements, production of defective items from a production line. In each case one
may have some knowledge of the likelihood of various possible results, but she/he cannot predict with any
certainty the outcome of any particular trial. Probability and statistics are used throughout engineering. Civil
engineers use statistics and probability to test and account for variations in materials and goods.

1.1.2. Some Important Terms

a. Probability is an area of study which involves predicting the relative likelihood of various outcomes. It is
a mathematical area which has developed over the past three or four centuries. Its usefulness for describing
errors of scientific and engineering measurements was soon realized. Engineers study probability for its
many practical uses, ranging from quality control and quality assurance to communication theory in
electrical engineering. Engineering measurements are often analyzed using statistics and a good knowledge
of probability is needed in order to understand statistics.
b. Statistics is a word with a variety of meanings. To the man in the street it most often means simply a
collection of numbers, such as the number of people living in a country or city, a stock exchange index, or
the rate of inflation. These all come under the heading of descriptive statistics. Another type of statistics will
engage people attention to a much greater extent. That is inferential statistics or statistical inference.
c. Chance is a necessary part of any process to be described by probability or statistics. Sometimes that
element of chance is due partly or even perhaps entirely to lack of knowledge of the details of the process.
For example, if one had complete knowledge of the composition of every part of the raw materials used to
make bolts, and of the physical processes and conditions in their manufacture, in principle she/he could
predict the diameter of each bolt. But in practice one generally lack that complete knowledge, so the
diameter of the next bolt to be produced is an unknown quantity described by a random variation. Under
these conditions the distribution of diameters can be described by probability and statistics. If one wants to
 2

improve the quality of those bolts and to make them more uniform, one will have to look into the causes
of the variation and make changes in the raw materials or the production process. But even after that, there
will very likely be a random variation in diameter that can be described statistically. Relations which involve
chance are called probabilistic or stochastic relations. These are contrasted with deterministic relations, in
which there is no element of chance. For example, Bernoulli’s Law and Newton’s Second Law involve no
element of chance, so they are deterministic.
d. Another term which requires some discussion is randomness. A random action cannot be predicted and so
is due to chance. A random sample is one in which every member of the population has an equal likelihood
of appearing. Just which items appear in the sample is determined completely by chance. If some items are
more likely to appear in the sample than others, then the sample is not random.

1.1.3. The Engineering Method and Statistical Thinking

An engineer is someone who solves problems of interest to society by the efficient application of scientific
principles. Engineers accomplish this by either refining an existing product or process or by designing a new
product or process that meets customers’ needs. The engineering, or scientific, method is the approach to
formulating and solving these problems. The steps in the engineering method are as follows:
1. Develop a clear and concise description of the problem.
2. Identify the important factors that affect this problem or that may play a role in its solution.
3. Propose a model for the problem, using scientific or engineering knowledge of the phenomenon
being studied. State any limitations or assumptions of the model.
4. Conduct appropriate experiments and collect data to test or validate the tentative model or
conclusions made in steps 2 and 3.
5. Refine the model on the basis of the observed data.
6. Manipulate the model to assist in developing a solution to the problem.
7. Conduct an appropriate experiment to confirm that the proposed solution to the problem is
both effective and efficient.
8. Draw conclusions or make recommendations based on the problem solution.
The engineering method steps are shown in Fig. 1.1. Steps 2–4 in Fig. 1.1 are enclosed in a box, indicating
that several cycles or iterations of these steps may be required to obtain the final solution. Consequently,
engineers must know how to efficiently plan experiments, collect data, analyze and interpret the data, and
understand how the observed data are related to the model they have proposed for the problem under study.

Figure 1.1. The Engineering Method Steps

The field of statistics deals with the collection, presentation, analysis, and use of data to make decisions,
solve problems, and design products and processes. Because many aspects of engineering practice involve
working with data, obviously some knowledge of statistics is important to any engineer. Specifically,
statistical techniques can be a powerful aid in designing new products and systems, improving existing
designs; and designing, developing, and improving production processes.
 3

Statistical methods are used to help people describe and understand variability. By variability, one mean
that successive observations of a system or phenomenon do not produce exactly the same result. Everybody
encounters variability in her/his everyday lives, and statistical thinking can give him/her a useful way to
incorporate this variability into decision-making processes. For example, consider the gasoline mileage
performance of a car. Do one always gets exactly the same mileage performance on every tank of fuel? Of
course not; in fact, sometimes the mileage performance varies considerably. This observed variability in
gasoline mileage depends on many factors, such as the type of driving that has occurred most recently (city
versus highway), the changes in condition of the vehicle over time (which could include factors such as tire
inflation, engine compression, or valve wear), the brand and/or octane number of the gasoline used, or
possibly even the weather conditions that have been recently experienced. These factors represent potential
sources of variability in the system. Statistics gives people a framework for describing this variability and
for learning about which potential sources of variability are the most important or which have the greatest
impact on the gasoline mileage performance.
A convenient way to think of a random variable, say X, that represents a measurement, is by using the model
X    (1.1)
where  is a constant and  is a random disturbance. The constant remains the same with every
measurement, but small changes in the environment, test equipment, differences in the individual parts
themselves, and so forth change the value of  . If there were no disturbances,  would always equal zero
and X would always be equal to the constant  . However, this never happens in the real world, so the actual
measurements X exhibit variability. One often needs to describe, quantify and ultimately reduce variability.

1.2. RELIABILITY ENGINEERING

Failures of major engineering systems always raise public concern on the safety and reliability of
engineering infrastructure. Decades ago quantitative evaluations of the reliability of complex infrastructure
systems were not practical, if not impossible. Engineers had to resort to the use of a safety factor mainly
determined through experience and judgment. Without exception, failures of hydrosystem infrastructure
(e.g., dams, levees, and storm sewers) could potentially pose significant threats to public safety and inflict
enormous damage on properties and the environment. The traditional approach of considering occurrence
frequency of heavy rainfalls or floods, along with an arbitrarily chosen safety factor, has been found
inadequate for assessing the reliability of hydrosystem infrastructure and for risk-based cost analysis and
decision making. In the past two decades or so, there has been a steady growth in the development and
application of reliability analysis in hydrosystems engineering and other disciplines.
Hydrosystems is the term used to describe collectively the technical areas of hydrology, hydraulics and
water resources. The term has now been widely used to encompass various water resource systems including
surface water storage, groundwater, water distribution, flood control, drainage, and others. In many
hydrosystem infrastructural engineering and management problems, both quantity and quality aspects of
water and other environmental issues have to be addressed simultaneously. Due to the presence of numerous
uncertainties, the ability of the system to achieve the goals of design and management decisions cannot be
assessed definitely. It is almost mandatory for an engineer involved in major hydrosystem infrastructural
design or hazardous waste management to quantify the potential risk of failure and the associated
consequences.
Occasionally, failures of engineering systems catch public attention and raise concern over the safety and
performance of the systems. The cause of the malfunction and failure could be natural phenomena, human
error, or deficiency in design and manufacture. Reliability engineering is a field developed in recent decades
to deal with such safety and performance issues.
 4

Based on their setup, engineering systems can be classified loosely into two types, namely, manufactured
systems and infrastructural systems. Manufactured systems are those equipment and assemblies, such as
pumping stations, cars, computers, airplanes, bulldozers, and tractors, that are designed, fabricated, operated,
and moved around totally by humans. Infrastructural systems are the structures or facilities, such as bridges,
buildings, dams, roads, levees, sewers, pipelines, power plants, and coastal and offshore structures, that are
built on, attached to, or associated with the ground or earth. Most civil, environmental, and agricultural
engineering systems belong to infrastructural systems, whereas the great majority of electronic, mechanical,
industrial, and aeronautical/aerospace engineering systems are manufactured systems.
The major causes of failure for these two types of systems are different. Failure of infrastructures usually is
caused by natural processes, such as geophysical extremes of earthquakes, tornadoes, hurricanes or
typhoons, heavy rain or snow, and floods, that are beyond human control. Failure of such infrastructural
systems seldom happens, but if a failure occurs, the consequences often are disastrous. Replacement after
failure, if feasible, usually involves so many changes and improvements that it is essentially a different, new
system. On the other hand, the major causes of failure for manufactured systems are wear and tear,
deterioration, and improper operation, which could be dealt with by human abilities but may not be
economically desirable. Their failures usually do not result in extended major calamity. If failed, they can be
repaired or replaced without affecting their service environment.
The performance of a hydrosystem engineering infrastructure, function of an engineering project, or
completion of an operation all involve a number of contributing components, and most of them, if not all,
are subject to various types of uncertainty (Fig. 1.2). Reliability and risk, on the other hand, generally are
associated with the system as a whole. Thus methods to account for the component uncertainties and to
combine them are required to yield the system reliability. Such methods usually involve the use of a logic
tree. The reliability of an engineering system may be considered casually, such as through the use of a
subjectively decided factor of safety. Today, reliability also may be handled in a more comprehensive and
systematic manner through the aid of probability theory.
The basic idea of reliability engineering is to determine the failure probability of an engineering system,
from which the safety of the system can be assessed or a rational decision can be made on the design,
operation, or forecasting of the system, as depicted in Fig. 1.3.
An infrastructure is a functioning system formed from a combination of a number of components. From the
perspective of reliability analysis, infrastructure systems can be classified in several ways. Infrastructures
may follow different paths to failure. The ideal and simplest type is the case that the resistance and loading
of the system are statistically independent of time, or a stationary system. Most of the existing reliability
analysis methods have been developed for such a case. A more complicated but realistic case is that for
which the statistical characteristics of the loading or resistance or both are changing with time, e.g., floods
from a watershed under urbanization, rainfall under the effect of global warming, sewer or water supply
pipes with deposition, and fatigue or elastic behavior of steel structure members. For some infrastructures,
the statistical characteristics of the system change with space or in time (or both), e.g., a reach of highway or
levee along different terrains.

1.3. DEFINITIONS OF RELIABILITY AND RISK

In view of the lack of generally accepted rigorous definitions for risk and reliability, it will be helpful to
define these two terms in a manner amenable to mathematical formulation for their quantitative evaluation
for engineering systems. Risk is defined as the probability of failure to achieve the intended goal. Reliability
is defined mathematically as the complement of the risk. In some disciplines, often the non-engineering
ones, the word risk refers not just to the probability of failure but also to the consequence of that failure,
such as the cost associated with the failure. Nevertheless, to avoid possible confusion, the mathematical
analysis of risk and reliability is termed herein reliability analysis.
 5

Figure 1.2. Sources of Uncertainties

Figure 1.3. Types of Reliability Engineering Problems

 6

Failure of an engineering system can be defined as a situation in which the load L (external forces or
demands) on the system exceeds the resistance R (strength, capacity, or supply) of the system. The
reliability ps of an engineering system is defined as the probability of non-failure in which the resistance of
the system exceeds the load; that is,

ps  P ( L  R ) (1.2a)

in which P (·) denotes probability. Conversely, the risk is the probability of failure when the load exceeds
the resistance. Thus the failure probability (risk) pf can be expressed mathematically as

p f  P( L  R)  1  ps (1.2b)

1.4. MEASURES OF RELIABILITY

In engineering design and analysis, loads usually arise from natural events, such as floods, storms, or
earthquakes, that occur randomly in time and in space. The conventional practice for measuring the
reliability of a hydrosystems engineering infrastructure is the return period or recurrence interval (T). The
return period is defined as the long-term average (or expected) time between two successive failure-causing
events. Simplistically, the return period is equal to the reciprocal of the probability of the occurrence of the
event in any one-time interval (T=1/p).
In fact, the conventional interpretation of return period can be generalized as the average time period or
mean time of the system failure when all uncertainties affecting load and resistance are considered. In other
words, the return period can be calculated as the reciprocal of the failure probability computed by Eq. (1.1b).
Two other types of reliability measures that consider the relative magnitudes of resistance and anticipated
load (called design load) are used frequently in engineering practice. One is the safety margin (SM), defined
as the difference between the resistance and the anticipated load, that is,

SM  R  L (1.3a)

The other is called the safety factor (SF), a ratio of resistance to load defined as

SF  R / L (1.3b)
 7

CHAPTER 2

BASIC CONCEPTS OF STATISTICS

2.1. INTRODUCTION

2.1.1. Definition of Statistics

Statistics can be defined in two senses: Plural (as Statistical Data) and singular (as Statistical Methods).
Plural Sense: Statistics are collection of facts (figures). This meaning of the word is widely used when
reference is made to facts and figures on sales, employment or unemployment, accident, weather, death,
education, etc. E.g.: Sales Statistics, Labor Statistics, Employment Statistics, etc. In this sense the word
Statistics serves simply as data. But not all numerical data are statistics.
Singular sense: Statistics is the science that deals with the methods of data collection, organization,
presentation, analysis and interpretation of data. It refers the subject area that is concerned with extracting
relevant information from available data with the aim to make sound decisions. According to this meaning,
statistics is concerned with the development and application of methods and techniques for collecting,
organizing, presenting, analyzing and interpreting statistical data.

2.1.2. Science of Statistics

Statistics is a branch of applied mathematics, which tries to solve the problems about random variables. The
goal of the statistics is to provide right (true) conclusions by using insufficient (deficient) data and
information. The events may be classified as deterministic (certain) and probabilistic (random) events. The
scope of the statistics science is only probabilistic events. A random or probabilistic event is described as, an
event which can not be certainly determined whether or not to occur, or if occurs, which value it will take.
Engineering problems have generally random characteristics. For example, it is impossible to certainly
predict that how much precipitation will take place next year in a location, this value may be predicted by
some probabilities (80, 90 or 99 percent).
The purpose of statistics is to develop and apply methodology for extracting useful knowledge from
experiments, measurements and data. Statisticians provide crucial guidance in determining what information
is reliable and which predictions can be trusted. There is a general perception that statistical knowledge is
frequently intentionally misused, by finding ways to interpret the data that are favorable to the presenter. A
famous quote is "There are three types of lies; lies, damn lies, and statistics."

2.1.3. Classification of Statistics

Based on the scope of the decision, statistics can be classified into two; Descriptive and Inferential Statistics.
Descriptive Statistics refers to the procedures used to organize and summarize masses of data. It is
concerned with describing or summarizing the most important features of the data. It deals only the
characteristics of the collected data without going beyond it. That is, this part deals with only describing the
data collected without going any further: that is without attempting to infer(conclude) anything that goes
beyond the data themselves. The methodology of descriptive statistics includes the methods of organizing
(classification, tabulation, Frequency Distributions) and presenting (Graphical and Diagrammatic
 8

Presentation) data and calculations of certain indicators of data like Measures of Central Tendency
and Measures of Dispersion (Variation) which summarize some important features of the data.
Inferential (Inductive) Statistics includes the methods used to find out something about a population, based
on the sample. It is concerned with drawing statistically valid conclusions about the characteristics of the
population based on information obtained from sample. In this form of statistical analysis, inferential
statistics is linked with probability theory in order to generalize the results of the sample to the population.
Performing hypothesis testing, determining relationships between variables and making predictions are also
inferential statistics.

2.1.4. Stages in Statistical Investigation

According to the singular sense definition of statistics, a statistical study (statistical investigation) involves
five stages: Collection, Organization, Presentation, Analysis and Interpretation of data.
a. Collection of Data: This is the first stage in any statistical investigation and involves the process of
obtaining (gathering) a set of related measurements or counts to meet predetermined objectives. The data
collected may be primary data (data collected directly by the investigator) or it may be secondary data (data
obtained from intermediate sources such as newspaper s, journals, official records, etc).
b. Organization of Data: It is usually not possible to derive any conclusion about the main features of the
data from direct inspection of the observations. The second purpose of statistics is describing the properties
of the data in a summary form. This stage of statistical investigation helps to have a clear understanding of
the information gathered and includes editing (correcting), classifying and tabulating the collected data in a
systematic manner. Thus the first step in the organization of data is editing. It means correcting (adjusting)
omissions, inconsistencies, irrelevant answers and wrong computations in the collected data. The second
step of the organization of data is classification that is arranging the collected data according to some
common characteristics. The last step of the organization of data is presenting the classified data in tabular
form, using rows and columns (tabulation).
c. Presentation of Data: The purpose of data presentation is to have an overview of what the data actually
looks like, and to facilitate statistical analysis. Data presentation can be done using Graphs and Diagrams
which have great memorizing effect and facilitates comparison.
d. Analysis of Data: The analysis of data is the extraction of summarized and comprehensive numerical
description in order to reach conclusions or provide answers to a problem. The problem may require simple
or sophisticated mathematical expressions.
e. Interpretation of Data: This is the last stage of statistical investigation. Interpretation involves drawing
conclusions from the data collected and analyzed in order to make decision.

2.1.5. Definition of Some Statistical Terms

A list of terms to be used in statistics are briefly defined as follows:

Population is a community, which is made of all of the components with a particular character. A population
is a totality of things, objects, peoples, etc about which information is being
Sample is a group of components which are supposed to represent the population. A sample is a subset or
part of a population selected to draw conclusions about the population.
Sampling is the process of selecting a sample from the population.
 9

Parameter is numerical values of population. Parameter is a descriptive measure (value) computed

from the population. It is the population measurement used to describe the population.
Statistics is numerical values of sample. It is a measure used to describe the sample; it is a value computed
from the sample. The parameter and statistic values are generally different. One of the main goals of
statistics science is to predict the statistic values as close as possible to parameter value.
Census Survey is the process of examining the entire population. It is the total count of the population.
Sampling Frame is a list of people, items or units from which the sample is taken.
Data are a collection of related facts and figures from which conclusions may be drawn.
Variable is a certain characteristic which changes from object to object and time to time.
Sample Size is the number of elements or observation to be included in the sample.
Collected is the totality of observations with which the researcher is concerned.

2.1.6. Applications, Uses and Limitations of Statistics

a. Applications of Statistics in Engineering:

In this modern time, statistical information plays a very important role in a wide range of fields. Today,
statistics is applied in almost all fields of human endeavor:
 In Scientific Research: Statistics is used as a tool in a scientific research. Statistical formulas and
concepts are applied on a data which are results of an experiment.
 In Quality Control: Statistical methods help to check whether a product satisfies a given standard.
 For Decision Making: Statistics helps to enhance the power of decision making in the face of
uncertainty by providing sufficient information.
 Reliability Engineering is the study of the ability of a system or component to perform its required
functions under stated conditions for a specified period of time.
 The application of probability theory, which includes mathematical tools for dealing with large
populations, to the field of mechanics, which is concerned with the motion of particles or objects
when subjected to a force.
 The field of statistics deals with the collection, presentation, analysis, and use of data to: Such as
make decisions, solve problems and design products and processes. It is the science of learning
information from data.
b. Uses of Statistics in Engineering:
 Design of Experiments (DOE) uses statistical techniques to test and construct models of engineering
components and systems.
 Quality control and process control use statistics as a tool to manage conformance to specifications
of manufacturing processes and their products.
 Time and methods engineering uses statistics to study repetitive operations in manufacturing in order
to set standards and find optimum (in some sense) manufacturing procedures.
 Reliability engineering uses statistics to measures the ability of a system to perform for its intended
function (and time) and has tools for improving performance.
 Probabilistic design uses statistics in the use of probability in product and system design.
 Every structural design, every safety factor, every hydrological analysis, every mechanical analysis,
everything, even the materials used are based on statistics. The results gotten from the analysis are
projected to other conditions, and the probability of them to interact together (for example,
earthquake, wind and max load or having the highest flow and rain).
 Condenses and summarizes masses of data and presents facts in numerical and definite form.
 10

 Facilitates comparison: Statistical devises such as averages, percentages, ratios, etc. are used for this
purpose.
 Formulating and testing hypothesis.
 Forecasting: Statistical methods help in studying past data and predicting future trends.
c. Limitations of Statistics:
 It cannot deal with a single observation; rather it deals aggregate of facts.
 Statistical methods are not applicable to qualitative character; it deals only with quantitative
characteristics.
 Statistical results are true on average; i.e. for the majority of case. Laws of statistics are not
universally true like the laws of physics, chemistry and mathematics.
 Statistics are liable to be misused or misinterpreted. This may be due to incomplete information,
inadequate and faulty procedures during data collection and sample selection and mainly due to
ignorance (lack of knowledge).

2.2. FREQUENCY ANALYSIS

2.2.1. Statistical Sample

The population, consisting of all the observations belonging to a random variable should be observed in
order to determine exactly the probability distribution of the random variables. However, in practise only a
statistical sample, which has finite number of elements, can be obtained from the population. A sample is a
set of observations or measurements collected to determine or estimate the statistical properties of a random
variable. Each element in the sample is an event, belonging to the random variable or is a value the random
variable has taken. Since the probability distribution function of the random variable and the parameters of
this distribution can only be estimated depending on the limited sample in hand, a sample must be analyzed
in an optimum way. Statistics is the science which obtains all possible information from samples and arrives
at conclusions about the statistical properties of the population, by using the obtained information. The
samples to be used in statistical studies should be adequate both qualitatively and quantitatively.
For qualitatively adequacy, a sample should realize the following conditions:
The data in the sample should be homogenous; in other words, all data should indeed be elements of the
population of the same random variable. Otherwise, the statistical calculations will have no significance. For
example, in the case the flow of a river is controlled by a dam; it would not be correct to evaluate the flows
downstream of the dam, measured before and after the dam construction, because these flows would not be
homogenous.
 There should be no systematic error in the measurement of the elements of the sample. In order to
meet this condition, the people responsible for sampling should be aware of both subject and
material; also, the measuring techniques to obtain the data should be adequate.
 Random errors should be minimized. This necessity can partly be realized by random (neutral)
sampling within the population.
The sample being quantitatively adequate means that, the number of elements in the sample is sufficiently
large. Although an exact limit can not be given for the sufficient number, it can easily be said that, as the
number of elements in the sample increase, more reliable results about the properties of the population can
be obtained. In statistics, samples having less than 25 to 30 elements are called small samples and in some
cases it is not suitable to use expressions that are valid for large samples in the analysis of small samples.
 11

2.2.2. Types of Variables

Variables can be classified as quantitative and qualitative variables. If a variable can be stated by a number,
it is a quantitative (numerical) variable. The other variables, which have no numerical character, are
qualitative variables. Since in statistics only quantitative variables are used, different classification as
discrete and continuous variables is made.
Discrete Variable: The variables of which number of components is limited (with small sample space) are
called as discrete variables. All of the qualitative variables are discrete. For example, the names of people,
colours of cars are discrete variables. The quantitative variables, which can be stated only by integer
numbers, in other words, the variables that are obtained by enumerating (1, 2, 3, ..) are also discrete
variables. For example, the numbers of students in a class or rainy days in a year are discrete variables.
Continuous Variable: If the number of components of a variable is unlimited or infinitive (with large sample
space), this variable is named as continuous. The quantitative variables that are obtained by measuring are
continuous and may have any fractional number (2.34, 145.036, …). Discharges of a stream, weights and
lengths of people are examples of continuous variables.

2.2.3. Frequency Analysis of Discrete Variables

In a sample with N component, if an X  xi event is occurred N i times, its frequency is

f ( xi )  f i  N i / N (2.1)
If these f ( xi ) frequency values are plotted in ordinate and x i values are shown in abscissa, then “frequency
graph or histogram” of the event is obtained. By adding or cumulating the frequency values and plotting
them versus x i values, then “cumulative frequency graph” is obtained. Cumulative frequencies are
calculated as follows:
i i
F ( xi )   N j / N   f ( x j ) (2.2)
j 1 j 1

2.2.4. Frequency Analysis of Continuous Variables

Since total number of a continuous random variable is theoretically infinitive (in other words, sample space
is very large), and total probability is equal to unity, the probability of a simple event is approximated to
zero (1/  0). So, the probability of combined events between (x ) and ( x  dx) interval is calculated,
instead of probability of a simple event, and a probability density function (p.d.f) is defined as (Figure 2.1a):

f ( x)dx  Px  X  x  dx (2.3)

The probability of X being between ( x1  x2 ) interval is calculated as

x2

P( x1  X  x2 )   f ( x)dx
x1
(2.4)

The cumulative distribution function of a real-valued random variable is the function given by (Figure 2.1b):
Fx ( x)  P( X  x) (2.5)

By differentiating this function, the probability density function (pdf) is obtained.

dFx ( x)
f x ( x)  (2.6)
dx
 12

a. Probability Density Functions b. Cumulative Frequency Distribution

Figure 2.1. Probability Density and Cumulative Frequency Distributions

Frequency analysis of a continuous random variable can be performed in two ways depending on the size of
the sample; whether it is a small or a large sample.

2.2.4.1. Frequency Analysis of Large Samples:

If the sample is large (at least have 25 to 30 samples), because of some difficulties in processing all of the
data, the range of the random variable is divided into appropriate number of class intervals. An important
point is to choose the number o class intervals (m). This number should be increased as the number of
elements in the sample increases. Generally, the number of class intervals is kept between 10 and 20. If too
few intervals are used in the analysis, a large amount of the information in the samples may be lost. On the
other hand, if too many intervals are used then both more effort than required will be needed and the
histogram will have a quite irregular shape, because very few or no observations will fall into some class
intervals. The following empirical formula can be used to determine the number of class intervals:

m  1  3.3 log10 N (2.7)

2.2.4.2. Frequency Analysis of Small Samples:
If the number of elements in the sample is small, it is not appropriate to classify the data. In this case, the
objective is to determine the cumulative frequency distribution only. The ordered sample is obtained by
listing the elements of the sample from smaller values to large values as:
x1  x 2  ...  x m  ...  x N (2.8)

One can calculate the frequency of the random variable remaining equal to or smaller than x m as:
P ( X  xm  F ( xm )  m / N (2.9a)
However, if this equation is applied, the frequency remaining equal to or smaller than the x N , the frequency
value is equal to 1. Since elements greater than the x N value may exist, it is not correct to use this
expression, which implies that the random variable (X) would never exceed the x N . Various formulas
called plotting position formulas have been proposed to eliminate this inconvenient aspect. The most
popular among them is as follows:
F ( x m )  m / N  1 (2.9b)

2.2.5. Parameters of Random Variables

The numbers, which express certain properties of the random variable’s distribution function, are called the
parameters of the distribution. Their estimation from the data is much easier compared to the estimation and
use of the distribution function.
 13

2.2.5.1. Central Parameters

a. The Mean: The most commonly used central parameter is called mean (expected value, arithmetic mean).
The mean in a population is calculated as:
  x/N (2.10a)

In a sample, the mean is calculated as:

x  x/ N (2.10b)
If the data are classified, then weighted mean is employed:

x   ( xi N i ) / N (2.10c)

b. The Mode: The mode is the most frequently occurring score value.

c. The Median: The median is the middle of a distribution: half the scores are above the median and half are
below the median.

d. The Geometric Mean: The geometric mean is the N th root of the product of the scores and is given as
follows:

G.M  N x1 * x 2 * ...... * x n (2.11)

e. The Harmonic Mean: Harmonic mean is used in calculating the mean slope of a highway, stream etc:

N
H .M .  (2.12)
1 / x1  1 / x2  ....  1 / x N

2.2.5.2. Variation Parameters

A deviation score is a measure of by how much each point in a frequency distribution lies above or below
the mean for the entire dataset as x  x , where x is raw score and x is the mean. A list of variation
parameters, which are used determine variations of all of the data around the centre of distribution, are
presented in the following:

a. The Range: The range is the difference between the maximum and minimum values of a series. Example:
The range of 35, 12, 34, 76, 87, 39, 48 is 87-12=75.
b. The Variance and Standard Deviation: The variance and the closely-related standard deviation are
measures of how spread out a distribution is. In other words, they are measures of variability. In order to
define the amount of deviation of a dataset from the mean, calculate the mean of all the deviation scores, i.e.
the variance. The variance is computed as the average squared deviation of each number from its mean. For
a continuous variance is defined as:

Varx   ( x  x )2 * f ( x)dx (2.13a)

and is calculated by a sample as:

Varx   ( xi  x ) 2 / N   xi / N  x 2
2
(2.13b)

If the number of data is less than 30 ( N  30) , in this equation ( N  1) is used instead of (N ) .If the data is
classified, the variance is calculated as:
 14


Varx   ( xi  x ) 2 * N i / N  (2.13c)

Since the variance has the square dimension of the variation, in order to make dimensional homogeneity, the
square root of variance, the standard deviation, is commonly used.

S x  Varx (2.13d)

c. Coefficient of Variance: It would not be proper to compare the standard deviations of two variables with
different averages in order to understand which variable has more variation; therefore, a dimensionless
coefficient, coefficient of variance is calculated as follows:

V .C.  S x / x (2.14)

2.2.5.3. Skewness Parameter

There are many different-shaped frequency distributions: J-shaped, Normal (Symmetrical), Rectangular,
Bimodal, Positive (Right) and Negative (Left) Skew (Figure 2.2).

Figure 2.2 Various Shaped Frequency Distributions

Skewness coefficient is employed to determine the shape of whether the distribution is symmetrical, right or
left skew. Skewness coefficient is positive, negative and zero for right and left skewed and symmetrical
distributions, respectively. Skewness coefficient is calculated as follows:

N2 m
Cs  * 13.5 (2.15a)
( N  1)( N  2) m2

Where, m2 is the second central moment (variance), m3 (the third central moment) is calculated:

m3   ( xi  x ) 3 / N (2.15b)

If the data are classified, m3 is found as:

m3   ( x i 
 x)3 Ni / N (2.15c)
 15

2.3. EXAMPLES

Example 2.1: Calculate the harmonic mean, standard deviation and skewness coefficient for the following
data. 25, 34, 43, 19, 36, 26, 38, 17, 25

Solution: 
x

xx

 x  x 2

x  x 3

Example 2.2: 32 yearly mean discharge values of a stream (m3/s) are given as follows. By taking class
interval as 3 m3/s, classify the data, obtain the frequency values and considering mid values of all classes,
calculate the mean and standard deviation:

28 19 16 11 19 20 17 15 13 16 24 13 18 20 23 20

15 13 10 17 21 19 18 24 12 21 25 26 13 18 27 14

Solution:

GROUP 
Ni

fi

xi

xi N i

xi  x 2 * N i
 16

Example 2.3: The following data are the mean annual precipitation heights (cm) of a city. Calculate
the mean, standard deviation and skewness coefficient by
a. Not classifying the data.
68 45 78 74 54 98 87 74 90 75 79 67 80 83 85 92 82 73 87 82 73 90 85 77 73 78 85 95 74 79 88 87
Solution:
Ni xi xi  x ( xi  x ) 2 ( xi  x )3 Ni xi xi  x ( xi  x ) 2 ( xi  x )3
1 68 17 82

2 45 18 73

3 78 19 87

4 74 20 82

5 54 21 73

6 98 22 90

7 87 23 85

8 74 24 77

9 90 25 73

10 75 26 78

11 79 27 85

12 67 28 95

13 80 29 74

14 83 30 79

15 85 31 88

16 92 32 87

TOTAL
 17

b. Classifying the data into 10 groups as (42-47), (48-53) ….. (98-103) and taking into consideration
of mid values of each group.
Solution:

Group Ni Mid. Val. xiNi ( xi  x) 2 * N i ( xi  x) 3 * N i

= xi
 18

c. Calculate and draw the frequency histogram for the classified data.
Solution:

Group

Ni

fi = Ni/N)*100


 19

CHAPTER 3

PROBABILITY

3.1. CONCEPT OF PROBABILITY

3.1.1. Definitions

Probability, as a specific term, is a measure of the likelihood that a particular event will occur. If one is
certain that an event will occur, its probability is 1 or 100%. If it certainly will not occur, its probability is
zero. The first situation corresponds to an event which occurs in every trial, whereas the second corresponds
to an event which never occurs. At this point one might be tempted to say that probability is given by
relative frequency, the fraction of all the trials in a particular experiment that give an outcome meeting the
stated requirements. But in general that would not be right. Why? Because the outcome of each trial is
determined by chance. Say one toss a fair coin, one which is just as likely to give heads as tails. It is entirely
possible that six tosses of the coin would give six heads or six tails, or anything in between, so the relative
frequency of heads would vary from zero to one. If it is just as likely that an event will occur as that it will
not occur, its true probability is 0.5 or 50%. But the experiment might well result in relative frequencies all
the way from zero to one. Then the relative frequency from a small number of trials gives a very unreliable
indication of probability. As an illustration, suppose the weather man on TV says that for a particular region
the probability of precipitation tomorrow is 40%. Let us consider 100 days which have the same set of
relevant conditions as prevailed at the time of the forecast. According to the prediction, precipitation the
next day would occur at any point in the region in about 40 of the 100 trials. (This is what the weather man
predicts, but we all know that the weather man is not always right!)
Although one cannot make an infinite number of trials, in practice she/he can make a moderate number of
trials, and that will give some useful information. The relative frequency of a particular event, or the
proportion of trials giving outcomes which meet certain requirements, will give an estimate of the
probability of that event. The larger the number of trials, the more reliable that estimate will be. Another
type of probability is the subjective estimate, based on a person’s experience. To illustrate this, say a
geological engineer examines extensive geological information on a particular property. He chooses the best
site to drill an oil well, and he states that on the basis of his previous experience he estimates that the
probability the well will be successful is 30%. (Another experienced geological engineer using the same
information might well come to a different estimate.) This, then, is a subjective estimate of probability. The
executives of the company can use this estimate to decide whether to drill the well.
Another approach is possible in certain cases. This includes various gambling games, such as tossing an
unbiased coin; drawing a colored ball from a number of balls, identical except for color, which are put into a
bag and thoroughly mixed; throwing an unbiased die; In each of these cases we can say before the trial that a
number of possible results are equally likely. This is the classical or a priori approach. The phrase “a priori”
comes from Latin words meaning coming from what was known before. This approach is often simple to
visualize, so giving a better understanding of probability.
 20

3.1.2. Basic Principles

Random variables cannot be studied by a deterministic approach; instead, a probabilistic approach is

required in their analysis. If a random variable is studied, the outcome of a future event can never be
determined with certainty. It is possible only to estimate the chance of the variable to assume a certain value.
The chance of the occurrence of a random variable is defined its probability. Total number of all of the
observations in an event or experiment is called total frequency.
The expected (anticipated) value of an event under normal conditions is named as expected frequency. The
probability of the event is determined as (expected frequency)/(total frequency). For example, when a coin is
tossed (thrown), under normal conditions the probabilities of head and tail are equal. If a coin is tossed 10
times, it is expected that 5 times head and 5 times tail to occur; the probability of either head or tail
is 5 / 10  0.5 . Denoting the random variable by a capital letter and its value in an observation by the
corresponding small letter, it can be written:
P( X  xi )  p i (3.1)
As it was previously explained, the frequency of an event is f ( xi )  f i  N i / N . The probability of this
event is defined as limit of its frequency as the number of observations approaches infinity:
pi  LimN i / N  (3.2)
N 
For example, the probability of head (or tail) is 0.5, which does not mean that in 10 tossing; certainly 5
heads and 5 tails will be observed. However, if the tossing is repeated more and more times, it is expected
that the numbers of heads and tails will approach to each other. In 1 000 000 tossing for example, it is highly
probable that nearly 500 000 heads and 500 000 tails will occur.
The basic axiom of the probability theory states that each random event has a certain probability that varies
in the range of 0 to 1. p i  0 implies that the event X  xi will never occur (impossible), pi  1 means
that the event will occur certainly (in all observations). For example, in a dice throwing, the probability of a
number between 1 and 6 is 1, and the probability of 0 or 7 is 0. The occurrence of an event is called as
success and non-occurrence is called as failure. If the probabilities of success and failure are displayed as
p and q , the equations can be written as:
p  q  1, p  1 q , q  1 p (3.3)
For example, the success and failure probabilities 2 in a dice are 1/6 and 1-1/6=5/6.

3.2. BASIC RULES OF COMBINING PROBABILITIES

3.2.1. Joint, Disjoint, Independent and Dependent Events

In the case of compound (combined) events, which have two or more simple events, if occurrence of one of
these events does not prevent the other(s) from occurring, these events are called as joint events. In other
words, if these events can occur simultaneously, they are joint events. Joint events have one or more than
one joint components. For example, in a dice tossing event, the events of (an odd number) and (a number
less than 4) are joint events; because they have joint (partner) components; as is seen, the numbers of 1 and
3 are joint components. If occurrence of one of these events prevents the other(s) from occurring, these
events are called as disjoint (mutually exclusive) events. In other words, if these events can not occur
simultaneously, they are disjoint events. Disjoint events have no joint component. For example, in a dice
tossing event, the events of (an odd number) and (an even number) are disjoint events; because they have
not joint (partner) components.
 21

In a compound event, if occurrence of an event does not affect the occurrence of the other events, these
events are called as independent events. For example, in the event of tossing two coins, the occurrence of tail
in each tossing does not affect the other and therefore these events are independent. If the occurrence of the
events affects each other, the events are dependent.

3.2.2. Addition Rule (“or”)

a. Disjoint (Mutually Exclusive) Events: In the case of two disjoint events, the occurrence probability of
(one or another), is the sum of their individually occurrence probabilities. The equation is:
P( A or B)  P( A  B)  P( A)  P( B) (3.4a)

For example, in a dice tossing event, the probability of (3 or 4) is

P(3  4)  P(3)  P(4)  1 / 6  1 / 6  1 / 3
In case of three or more events, the same rule is valid. In three disjoint events the rule is:
P( A or B or C )  P ( A  B  C )  P ( A)  P ( B)  P (C ) (3.4b)

For example, the probability of even number in a dice is

P(even)  P(2)  P(4)  P(6)  1 / 6  1 / 6  1 / 6  1 / 2

b. Joint Events: In the case of two joint events, the occurrence probability of (one or another), is obtained by
subtracting the probability of joint event from the sum of their individually occurrence probabilities. In
Figure 3.2b, the following equation may be obtained:
P( A or B)  P( A  B)  P( A)  P( B)  P( A  B) (3.5a)

For example: In a dice, calculate the probability of (an odd number) or (a number less than 3)
Solution: P( odd number )  P( A)  P(1)  P(3)  P(5)  1 / 6  1 / 6  1 / 6  1 / 2,
( P less than 3 )  P( B)  P(1)  P(2)  1 / 6  1 / 6  1 / 3 , P ( A  B )  P (1)  1 / 6
From Equation (2.18a) P( A or B)  1 / 2  1 / 3  1 / 6  2 / 3

In the case of three joint events, the probability is calculated as follows:

P( A  B  C )  P( A)  P( B)  P(C )  P( A  B)  P( A  C )  P ( B  C )  P( A  B  C ) (3.5b)

3.2.3. Multiplication Rule (“and”)

a. Independent Events: The probability of independent events is equal to multiplication of their individual
probabilities. In other words; the probability of occurrence of more than one event together is the product of
the probabilities of the separate events.
P( A and B )  P( A  B )  P( A) * P ( B) (3.6a)

P( A and B and C )  P( A  B  C )  P( A) * P( B) * P(C ) (3.6b)

For example, if a dice is thrown twice, the probability of 6 in both event is P(6,6)  1 / 6 * 1 / 6  1 / 36 .
In the case of independent events more than 3, the same rule is valid.
 22

b. Dependent Events: If the events are not independent, one event affects the probability for the other event. In
this case conditional probability must be used. The conditional probability of B given that A occurs, or on
condition that A occurs, is written P(B/A). This is read as the probability of B given A, or the probability of
B on condition that A occurs. Conditional probability can be found by considering only those events which
meet the condition, which in this case is that A occurs. Among these events, the probability that B occurs is
given by the conditional probability, P(B/A).
The multiplication rule for the occurrence of both A and B together when they are not independent is the
product of the probability of one event and the conditional probability of the other:
P( A  B)  P( A) * P( B / A)  P( B) * P( A / B) (3.7a)
This implies that conditional probability can be obtained by:

P( A  B) P( A  B)
P( B / A)  , P( A / B)  (3.7b)
P( A) P( B)

Example: Knowing that when a dice is tossed, the number on top is greater than 2, what is the probability
that the number on the top of the dice is an even number?
Solution: The event that number is greater than 2 is A and the event of even number is B, then
A is (3, 4, 5, 6) number of elements of A is 4 and P(A) = 4/6,
B is (2, 4, 6), number of elements of B is 2 and P(B) = 2/6,
Event of (A  B) = (4, 6), number of elements is 2 and P(A  B) = 2/6
Then, the probability is
P( A  B) 2 / 6 1
P( B / A)    is found.
P( A) 4/6 2
Addition and multiplication rules are summarized in Table 3.1.
Table 3.1. Addition and Multiplication Rules
Rule Type of Events
Addition (“or’) Disjoint Events Joint Events
P( A  B)  P( A)  P( B) P( A  B)  P( A)  P ( B)  P( A  B )

Multiplication (“and”) Independent Events Dependent Events

P( A  B)  P( A) * P( B) P( A  B) P( A  B)
P( B / A)  , P( A / B) 
P( A) P( B)
 23

3.3. EXAMPLES

Example 3.1: The probabilities of earthquake, material failure and foundation failure of a building are 3, 6,
and 8 percent, respectively. Assuming that these three events are independent:

a. Calculate the probability of (earthquake or foundation failure)

b. The building will be destroyed (demolished) even if one of these three events is occurred; calculate the
probability of the building to be destroyed.

Solution:
 24

Example 3.2: The probabilities of flood (F), earthquake (E), and hurricane (H) in a region are 10%, 5%, and
8%, respectively. Calculate the probabilities of:
a. No event occurrence, b. (F or E) and (F or E or H).
Solution:

Example 3.3. The water of a city is transmitted by A, B and C pipes. The discharges of the pipes are
QA = 10 l/s, QB = 15 l/s and QC = 25 l/s. The failure probabilities of these pipes are PA=0.02, PB=0.04 and
PC=0.06 and their failures are independent. Calculate the probabilities for the city:

a. Transmission of all of the water, b. Being the transmitted discharge at least 40 l/s, c. Being without water.

Solution:
 25

CHAPTER 4

IMPORTANT PROBABILITY FUNCTIONS

4.1. PROBABILITY DISTRIBUTIONS

The behavior of a random variable is characterized by its probability distribution, that is, by the way
probabilities are distributed over the values it assumes. A probability distribution function and a probability
mass function are two ways to characterize this distribution for a discrete random variable. They are
equivalent in the sense that the knowledge of either one completely specifies the random variable. The
corresponding functions for a continuous random variable are the probability distribution function, defined
in the same way as in the case of a discrete random variable, and the probability density function.
Given a random experiment with its associated random variable and given a real number, x let us consider
the probability of the event P ( X  x) . This probability is clearly dependent on the assigned value x. The
following function is defined as the Probability Distribution Function (PDF), or simply the distribution
function of X.

FX ( x)  P( X  x) (4.1)

In Equation (4.1), subscript X identifies the random variable. This subscript is sometimes omitted when
there is no risk of confusion. Let us repeat that FX (x) is simply P(A), the probability of an event A
occurring, the event being P ( X  x) .
The PDF is thus the probability that X will assume a value lying in a subset of S, the subset being point x
and all points lying to the ‘left’ of x. As x increases, the subset covers more of the real line, and the value of
PDF increases until it reaches 1. The PDF of a random variable thus accumulates probability as x increases,
and the name Cumulative Distribution Function (CDF) is also used for this function.
In view of the definition and the discussion above, one gives below some of the important properties
possessed by a PDF.
 It exists for discrete and continuous random variables and has values between .0 and 1.
 It is a nonnegative, continuous-to-the-left, and nondecreasing function of the real variable Moreover,
one has

FX ()  0 and FX ()  1 (4.2)

 If and are two real numbers such that a<b, then

P(a  X  b)  FX (b)  FX (a) (4.3a)

This relation is a direct result of the identity

P ( X  b)  P ( X  a )  P ( a  X  b) (4.3b)
 26

One can see from Eq. (4.3a) that the probability of X having a value in an arbitrary interval can be
represented by the difference between two values of the PDF. Generalizing, probabilities associated with
any sets of intervals are derivable from the PDF.
Example: Let a discrete random variable X assume -1, 1, 2, and 3 with probabilities 1/4, 1/8, 1/8 and 1/2,
respectively. One then has

This function is plotted in Fig. (4.1a). It is typical of PDFs associated with discrete random variables,
increasing from 0 to 1 in a ‘staircase’ fashion.
A continuous random variable assumes a none numerable number of values over the real line. Hence, the
probability of a continuous random variable assuming any particular value is zero and therefore no discrete
jumps are possible for its PDF. A typical PDF for continuous random variables is shown in Fig. (4.1b). It
has no jumps or discontinuities as in the case of the discrete random variable. The probability of X having a
value in a given interval is found by using Eq. (4.3a), and it makes sense to speak only of this kind of
probability for continuous random variables. For example, in Fig. (4.2b)

P(1  X  1)  FX (1)  FX (1)  0.8  0.4  0.4

Clearly, P(x=a) = 0, for any a.

a. Random Variable b. Continuous Variable

Figure 4.1. Probability Distribution Functions of X, FX(x) for Discrete and Random Variables
For a continuous random variable X, its PDF, FX(x) is a continuous function of x and the derivative

dFX ( x)
f X ( x)  (4.4)
dx

exists for all x. The function is called the probability density function (pdf), or simply the density function of
X. (Note the use of upper-case and lower-case letters, PDF and pdf, to represent the distribution and density functions,
respectively).
Since is monotone nondecreasing, one clearly has for all x:
f X ( x)  0 (4.5)
 27

Additional properties of can be derived easily from Eq. (4.4); these include


f

X ( x)dx  1 (4.6a)

and
b
P(a  X  b)  FX (b)  FX (a)   f X ( x)dx (4.6b)
a
An example of pdfs has the shape shown in Fig. (4.2). As indicated by Eq. (4.6a and 4.6b), the total area
under the curve is unity and the shaded area from to gives the probability P(a  X  b) . One again observes
that the knowledge of either pdf or PDF completely characterizes a continuous random variable. The pdf
does not exist for a discrete random variable since its associated PDF has discrete jumps and is not
differentiable at these points of discontinuity.

Figure 4.2. A Probability Density Function, fX(x)

Using the mass distribution analogy, the pdf of a continuous random variable plays exactly the same role as
the probability mass function (pmf) of a discrete random variable. The function fX(x)can be interpreted as
the mass density (mass per unit length). There are no masses attached to discrete points as in the discrete
random variable case. The use of the term is density function therefore appropriate here for fX(x).
Many of the probability problems encountered in application fit some known probability functions. Since
the features of these functions are known, various problems can be easily solved by using them. Probability
functions in engineering practice can be classified as related to discrete variables and continuous variables.

4.2. FUNCTIONS OF DISCRETE VARIABLES

4.2.1. Binom Distribution

The events that have two alternatives are commonly encountered in practice. One of these alternatives is
success and the other is fail, of which probabilities are p and q, respectively (p+q=1). For example, in a coin
tossing event, the probabilities of tail and head are equal to 0.5. Whether or not a flood will be encountered
in a year is another example. Let a sample with N component is taken from these variables; in other words,
let N independent trials are performed. These are called Bernoulli Trials. The probability of x successes (x
an integer between 0 and N) of an event, of which probability is p, in N trials fits Binom Distribution and
can be computed as:

N!
P( x)  p x q N x (4.7a)
x!( N  x)!
Binom Distribution has the following parameters:

Expected value = Mean : E x  x  Np , Variance: V x  Npq (4.7b)

 28

4.2.2. Poisson Distribution

Poisson distribution is a limit case of Binom and is successfully applied for calculating the probabilities of
seldom (with have very small probability) simple events (heavy rain, flood, hurricane etc). The probability
of x successes in N trials is calculated by:

x
P( X  x)  (4.8a)
e  x!

Poisson distribution has one parameter ()

E x  x  V x    Np (4.8b)

4.3. FUNCTIONS OF CONTINUOUS VARIABLES

4.3.1. Normal Distribution

A number of random variables encountered in practical applications fit to the normal (Gaussian) distribution
with the following probability density function (pdf):

1  1  x   2 
P( x )  exp     (4.9)
 2  2   

It has two parameters; , mean and , standard deviation; it is symmetrical (Cs=0).Total area covered by
normal curve is equal to 1. The probability of x between ( x1 , x 2 ) interval is equal to the area between these
values. The probabilities of the normal variable to remain in the intervals around the mean one, two and
three standard deviations are equal to 0.6826, 0.9544 and 0.9974 (  1), respectively (Figure 4.3).

P(x)

P(-<x<+)=0.6826

P(x1<x<x2)
P(x<a) P(x>b)

x
a x1 x2 -  + b

Figure 4.3. Properties of Normal Distribution

 29


P(  x  )   P( x)dx  1

 
P( x   )  P( x   )   P( x)dx   P( x)dx  0.5
 
a 
(4.10)
x2
P( x  a) 

 P( x)dx, P( x  b)   P( x)dx, P( x
b
1  x  x2 )   P( x)dx
x1

 3
P(   3  x    3 )   P( x)dx  0.9974  1
 3

The probability distribution function F (x ) of the normal distribution is tabulated numerically, by

standardizing the random variable as follows:

x xx
z  (4.11)
 Sx

The distribution of z is called as standard normal distribution. All of the rules of normal distribution are also
valid for standard normal distribution. The probability density function of standard normal distribution is:
1  z2 /2
P( z)  e (4.12)
2
The tabulated values of normal distribution can be used for calculating the probability of z is between zero
and a positive z1 (Table 4.1). Since the distribution is symmetrical, the same probabilities are valid for
negative z values. The probability can be calculated by the following equation:

 
  83z  351z  562 
p  0.5  0.5 * exp  (4.13a)
 703
 165 
 
 z 

Similarly, the related z value for p probability is calculated as:

z
0.5  p 
0.135
 0.5  p 
0.135
(4.13b)
0.1975

4.3.2. Lognormal Distribution

All of the variables do not fit to normal distribution; however, it is often attempted to transform a non-
normal random variable because the normal distribution has well known properties and is easy to use. The
most commonly used transformation is the logarithmic transformation. If the transformed variable

Y  ln X (4.14)

fits to the normal distribution, then the distribution of the original variable X is called lognormal. The
distribution is defined only for positive values.
 30

Table 4.1. Standard Normal Distribution

z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.0 0.0000 0.0040 0.0080 0.0120 0.0160 0.0190 0.0239 0.0279 0.0319 0.0359
0.1 0.0398 0.0438 0.0478 0.0517 0.0557 0.0596 0.0636 0.0675 0.0714 0.0753
0.2 0.0793 0.0832 0.0871 0.0910 0.0948 0.0987 0.1026 0.1064 0.1103 0.1141
0.3 0.1179 0.1217 0.1255 0.1293 0.1331 0.1368 0.1406 0.1443 0.1480 0.1517
0.4 0.1554 0.1591 0.1628 0.1664 0.1700 0.1736 0.1772 0.1808 0.1844 0.1879
0.5 0.1915 0.1950 0.1985 0.2019 0.2054 0.2088 0.2123 0.2157 0.2190 0.2224
0.6 0.2257 0.2291 0.2324 0.2357 0.2389 0.2422 0.2454 0.2486 0.2517 0.2549
0.7 0.2580 0.2611 0.2642 0.2673 0.2704 0.2734 0.2764 0.2794 0.2823 0.2852
0.8 0.2881 0.2910 0.2939 0.2969 0.2995 0.3023 0.3051 0.3078 0.3106 0.3133
0.9 0.3159 0.3186 0.3212 0.3238 0.3264 0.3289 0.3315 0.3340 0.3365 0.3389
1.0 0.3413 0.3438 0.3461 0.3485 0.3508 0.3513 0.3554 0.3577 0.3529 0.3621
1.1 0.3643 0.3665 0.3686 0.3708 0.3729 0.3749 0.3770 0.3790 0.3810 0.3830
1.2 0.3849 0.3869 0.3888 0.3907 0.3925 0.3944 0.3962 0.3980 0.3997 0.4015
1.3 0.4032 0.4049 0.4066 0.4082 0.4099 0.4115 0.4131 0.4147 0.4162 0.4177
1.4 0.4192 0.4207 0.4222 0.4236 0.4251 0.4265 0.4279 0.4292 0.4306 0.4319
1.5 0.4332 0.4345 0.4357 0.4370 0.4382 0.4394 0.4406 0.4418 0.4429 0.4441
1.6 0.4452 0.4463 0.4474 0.4484 0.4495 0.4505 0.4515 0.4525 0.4535 0.4545
1.7 0.4554 0.4564 0.4573 0.4582 0.4591 0.4599 0.4608 0.4616 0.4625 0.4633
1.8 0.4641 0.4649 0.4656 0.4664 0.4671 0.4678 0.4686 0.4693 0.4699 0.4706
1.9 0.4713 0.4719 0.4726 0.4732 0.4738 0.4744 0.4750 0.4756 0.4761 0.4767
2.0 0.4772 0.4778 0.4783 0.4788 0.4793 0.4798 0.4803 0.4808 0.4812 0.4817
2.1 0.4821 0.4826 0.4830 0.4834 0.4838 0.4842 0.4846 0.4850 0.4854 0.4857
2.2 0.4861 0.4864 0.4868 0.4871 0.4875 0.4878 0.4881 0.4884 0.4887 0.4890
2.3 0.4893 0.4896 0.4898 0.4901 0.4904 0.4906 0.4909 0.4911 0.4913 0.4916
2.4 0.4918 0.4920 0.4922 0.4925 0.4927 0.4929 0.4931 0.4932 0.4934 0.4936
2.5 0.4938 0.4940 0.4941 0.4943 0.4945 0.4946 0.4948 0.4949 0.4951 0.4952
2.6 0.4953 0.4955 0.4956 0.4957 0.4959 0.4960 0.4961 0.4962 0.4963 0.4964
2.7 0.4965 0.4966 0.4967 0.4968 0.4969 0.4970 0.4971 0.4972 0.4973 0.4974
2.8 0.4974 0.4975 0.4976 0.4977 0.4977 0.4978 0.4979 0.4979 0.4980 0.4981
2.9 0.4981 0.4982 0.4982 0.4983 0.4984 0.4984 0.4985 0.4985 0.4986 0.4986
3.0 0.4987 0.4987 0.4987 0.4988 0.4988 0.4989 0.4989 0.4989 0.4990 0.4990
3.1 0.4990 0.4991 0.4991 0.4991 0.4992 0.4992 0.4992 0.4992 0.4993 0.4993
3.2 0.4993 0.4993 0.4994 0.4994 0.4994 0.4994 0.4994 0.4995 0.4995 0.4995
3.3 0.4995 0.4995 0.4995 0.4996 0.4996 0.4996 0.4996 0.4996 0.4996 0.4997
3.4 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4998
 31

4.3.3. Gumbel Distribution

Gumbel (extreme value type I) distribution has two forms. One is based on the smallest extreme and the
other is based on the largest extreme. The general formula for the probability density function of the
maximum Gumbel distribution is
1  x x0  e  x  x0 
f ( xi )  e e (4.15)
a
where x0 is the location parameter and a is the scale parameter. The formula for the cumulative distribution
function of the Gumbel distribution is
 a ( x x 0 )
F ( xi )  P( x  xi )  1  e e (4.16)
SN S
a , x0  x  y N x (4.17)
Sx SN
Let y  a( x  x0 ) , then one can obtain,
y y
F ( xi )  P ( x  xi )  1  e  e  1  F ( x i )  P ( x  xi )  e  e (4.18a)
From this equation, one can obtain
y   ln ln1  F ( xi ) (4.18b)
For example, F ( xi )  0.01  y   ln ln1  0.01  4.600 , F ( x)  0.1  y   ln ln1  0.1  2.250

SN  Sx 
y value is calculated as y  a ( x  x0 )  x  (x  yN 
Sx  SN 

SN   S  S  y  S y
yi   xi   x  y N x   N  xi  x  S x N   xi  yi x  x  S x N (4.19)
Sx   S N  S x  SN  SN SN

Where, x and S x are the mean and standard deviation of the data, respectively. The values of S N and y N are
tabulated as follows according to the number of sample (N) and are given in Table 4.2 .
Table 4.2. yN and SN Values for Gumbel Distribution
N 10 15 20 25 30 35 40 50 75 100 200
yN 0.495 0.513 0.524 0.531 0.536 0.540 0.544 0.549 0.556 0.560 0.567
SN 0.950 1.021 1.063 1.092 1.112 1.129 1.141 1.161 1.190 1.207 1.236

4.3.4. Pearson Type III Distribution

The x value, of which exceedance probability is p is calculated as

3
2  zCs Cs 2  2
x  x  KS x , K  1     (4.20)
Cs  
 6 36  Cs

Where, Cs is the skewness coefficient and z is the standard normal value.

If y  ln x values fit Pearson Type III Distribution, then the distribution of the original variable X is called
Log Pearson Type III Distribution.

Log Normal (LN,) Gumbel (G), Pearson (P III) and Log-Pearson (LP III) Type III Distributions are often
used in extreme values (floods, droughts, heavy rains etc.).
 32

4.4. EXAMPLES

Example 4.1: The spillway of a dam is designed according to a flood with annual probability is 0.33
percent. Calculate the non-occurring probability and 1 times and 2 times occurring probabilities of this flood
in 45 years according to both Binom and Poisson Distributions.
Solution:
Binom Distribution:
p = 0.0033, q = 1 – 0.0033 = 0.9967, N=45
N! 45!
Eq. (4.7a): P( x)  p x q N x  * 0.0033x * 0.996745 x
x!( N  x)! x!(45  x)!
45!
Non-occurring = 0 times occurring x = 0  P(0)  * 0.00330 * 0.9967450  0.8618
0!(45  0)!
45!
x = 1 times occurring  P(1)  * 0.00331 * 0.9967451  0.1283
1!(45  1)!
45!
x = 2 times occurring  P(2)  * 0.00332 * 0.9967452  0.00935
2!(45  2)!

Poisson Distribution:
x x 0.1485x
Eq. (4.8a): P ( x)  , Eq. (4.8b):   Np  45 * 0.0033  0.1485 , P( x)  
e  x! e  x! e 0.1485 * x!
0.14850
Non-occurring = 0 times occurring x = 0  P(0)   0.8620
e 0.1485 * 0!
0.14851
x = 1 times occurring  P(1)   0.1280
e 0.1485 *1!
0.14852
x = 2 times occurring  P(2)  0.1485  0.0095
e * 2!

Example 4.2: Calculate the probability of a traffic accident occurring 0 times and 5 times in 50 years,
according to Binomial and Poisson distributions, with a 20% probability to occur in any year.

Solution:

Binom Distribution:
50! 50!
P(0)  0.2 0 * 0.850  1.427 *105 , P(5)  0.2 5 * 0.8 45  0.0295
0!(50  0)! 5!(50  5)!

Poisson Distribution:
0.101 105
  50 * 0.20  10 ,  P(0)   4. 54 * 10 5
, , P(5)  10  0.0378
e10 *1! e 5!
 33

Example 4.3: Calculate the probability of z variable is between -2 and 2.

Solution:

Example 4.4: Calculate of area of between x = 4 and x = 9 in a Normal Distribution with a mean of 5 and
standard deviation of 2. In other worlds, calculate the probability of the value is between 4 and 9.
Solution:

Example 4.5: Total annual precipitation height of a gauge station fit Normal Distribution with a mean of 90
cm and standard deviation of 25 cm. Calculate the probabilities of total annual precipitation in any year are:
a. Less than 70 cm, b. Between 80 and 105 cm, c. Between 60 and 130 cm, d. More than 140 cm

Solution:
 34

Example 4.6: Mean and standard deviation values of total annual precipitation heights of a city are 60 cm
and 15 cm, respectively. Assuming that the data fit Normal Distribution:
a. Determine the interval which contains 95 percent of the data (95% confidence interval),
b. A year with less than 45 cm precipitation height is called as “drought year” and with greater than 90 cm
precipitation height is called as “flood year”. Calculate the expected numbers of both drought and flood
years in 50 years.
c. Estimate the maximum and minimum annual precipitation heights in 50 years.

Solution:

b.
 35

c.

Example 4.7: The mean annual discharge values of a stream fit Normal Distribution with a mean of 70 m 3/s

and a standard deviation of 10 m3/s. With the 80 yearly duration:

a. Estimate the number of years with less than 78 m3/s mean discharge,
b. Calculate the 92 percent confidence interval of mean of the population.

Solution:
 36

Example 4.8: Yearly maximum discharges of a stream (m3/s) are given as follows. Calculate the probability
that discharges are between (55 and 70) m3/s, greater than 75 m3/s; the discharge values of which
exceedance probabilities are 1 and 0.1 percent; by using the distributions of: a. Gumbel (G), b. Log Normal
(LN), c. Log Pearson Type III (LPT)
68 76 39 48 57 71 63 56 37 54 59 70 54 62 68

Solution:
 37

CHAPTER 5

SAMPLING DISTRIBUTIONS

5.1. THE CONCEPT OF SAMPLING DISTRIBUTION

In chapter 2.2.5, several summary statistics were presented which described key attributes of a dataset. They
were sample estimates (such as x and S x ) of true and unknown population parameters (such as μ and σ). In
this chapter, descriptions of the uncertainty or reliability of sample estimates are presented. As an alternative
to reporting a single estimate, the utility of reporting a range of values called an interval estimate is
demonstrated. The sample mean and standard deviation estimate the corresponding points of a population.
Such estimates are called point estimates. By themselves, point estimates do not portray the reliability, or
lack of reliability (variability), of these estimates. For example, suppose that two data sets X and Y exist,
both with a sample mean of 5 and containing the same number of data. The Y data all cluster tightly around
5, while the X data are much more variable. The point estimate of 5 for X is much less reliable than that for
Y because of the greater variability in the X data. In other words, more caution is needed when stating that 5
estimates the true population mean of X than when stating this for Y. Reporting only the point estimate of 5
fails to give any hint of this difference.
As an alternative to point estimates, interval estimates are intervals which have a stated probability of
containing the true population value. The intervals are wider for data sets having greater variability. Thus in
the above example an interval between 4.7 and 5.3 may have a 95% probability of containing the (unknown)
true population mean of Y. It would take a much wider interval, say between 2.0 and 8.0, to have the same
probability of containing the true mean of X. The difference in the reliability of the two estimates is
therefore clearly stated using interval estimates. Interval estimates can provide a piece of information which
point estimates cannot: A statement of the probability that the interval contains the true population value (its
reliability).
The real value of any  parameter of a random variable is never certainly determined, because it is
impossible to observe the whole population of this variable. However, it is possible to calculate a value of
b statistics which is an estimate of this parameter from a sample, to be supposed to represent the population.
Statistics b is highly probably not equal the parameter  , it is the best estimate of  which can be obtained
from the sample at hand. The calculated b statistics (for example the mean values, x i ) from various samples
drawn from the same population are only estimates of the corresponding parameter (for example mean value
of the population, ). The values of a statistics calculated from different samples have a distribution, because
any statistic value can be treated as a random variable. The probability distribution of the values of any
statistic to be calculated of same population with same size is called the sampling distribution of this
statistics.
 38

To know the sampling distribution of a statistic is important for the following reason: As mentioned
above, the statistics determined from the sample at hand is not equal to the real value of the population
parameter. Without observing the whole population, it is impossible to determine the value of parameter
with an absolute correctness. However, it can be determined the interval in which the unknown parameter
will remain around the calculated statistic with a given probability. For this, the sampling distribution of that
statistics must be known.
The sampling distribution f (b) of the b statistics of the  parameter calculated from samples of magnitude
N is presented in Figure 5.1. The expected value of this distribution E (b) will be the b0 value calculated from
the sample at hand. For determining (b1 , b2 ) interval in which the unknown  parameter will remain with a
given Pc probability, such a symmetrical interval (b1 , b2 ) is chosen around b0 that, the percentage of the
sampling distribution within this interval is Pc . Here, Pc is the probability that the interval includes the true
value and is called the confidence level and the interval (b1 , b2 ) is named the confidence interval at this
confidence level and   1  Pc is the probability that this interval will not cover the true value and is called
the significance level. Values as 0.90, 0.95 and 0.99 are used for Pc in practice.

f(b)

Pc =1- (confidence level)

/2 (1-Pc)/2=/2

b1 E(b)=b0 b2 b
Confidence Interval

Figure 5.1. Confidence Level and Confidence Interval

As the number of the elements of the sample (N ) increases, the confidence interval corresponding to a
certain confidence level gets narrower. In other words, the confidence interval within which the parameter
will remain with a certain probability is smaller for large samples, expressing that the error in parameter
estimation is reduced. In order to estimate the parameter of the population, the sampling distribution related
to the parameter is used. Sampling distributions are determined theoretically for some statistics.

5.2. ESTIMATION OF CONFIDENCE INTERVALS OF POPULATION PARAMETERS

In the estimation of population parameters with a confidence level, two different methods are used according
to asymptotic and exact distributions. In general, it is adequate to estimate the confidence intervals of only
two important parameters, the mean and the variance (or the standard deviation).
 39

5.2.1. Estimation of Confidence Interval of Mean

a. In the Event of Asymptotic Distribution:

If the number of elements is large ( N  30), sampling distribution of means that are calculated from
different samples fit to Normal Distribution. Accordingly, this sampling distribution has a mean, a standard
deviation; moreover, its values within a confidence interval can be estimated by Normal Distribution. The
confidence interval of the mean of sampling distribution’s mean is calculated as follows:

Pc  P(b1    b2 )  Px  zS x     x  zS x  (5.1)

This equation implies that, the mean of population (  ) will be within the (b1 , b2 ) interval with Pc
probability; in other words, the confidence interval with Pc probability of  is (b1 , b2 ) . In Equation 5.1, the
z value is obtained from Normal Distribution Table corresponding the value of
0.5   / 2  0.5  1  Pc  / 2  Pc / 2 . For example, the z value for 0.95 confidence level is obtained as to
correspond the value of 0.95 / 2  0.475  z  1.96 . The S x value in (5.1) the standard deviation of
sampling distribution of x variable and is calculated as follows:

Sx
Sx  (5.2a)
N
If this value is inserted in (5.1), the following equation is obtained:

 
Pc  P(b1    b2 )  P x  zS x / N    x  zS x / N   (5.2b)

In other words, the lower and upper limits of confidence interval of the mean the population (  ) are:

Sx Sx
Lower limit: b1  x  z , upper limit: b2  x  z (5.2c)
N N
b. In the Event of Exact Distribution:
If the amplitude of the data is small ( N  30) , the confidence interval of the mean is calculated by t
(student) distribution as follows:

xx
t
Sx /  N 1  (5.3)

The t distribution is symmetric and the probability of t is greater than a given value of t 0 is tabulated
according to ( N  1) degree of freedom for various confidence levels (Table 5.1). For large N values,
t distribution approaches Normal Distribution. The confidence interval of parameter  is given as:

 
Pc  P(b1    b2 )  P x  tS x / N  1    x  tS x / N  1   (5.4a)

In other words, the lower and upper limits of confidence interval of the mean of the population (  ) are:

Sx Sx
Lower limit: b1  x  t , upper limit: b2  x  t (5.4b)
N 1 N 1
As is seen, when this equation is compared to (5.2c), t and N  1 values have come instead of z and N ,
respectively.
 40

TABLE 5.1. t (STUDENT) DISTRIBUTION (df: DEGREE OF FREEDOM)

Pc Pc

df 0.90 0.95 0.98 0.99 df 0.90 0.95 0.98 0.99

1 6.314 12.706 31.821 63.657 2 2.920 4.303 6.965 9.925
3 2.353 3.182 4.541 5.841 4 2.132 2.776 3.747 4.604
5 2.015 2.571 3.365 4.032 6 1.943 2.447 3.143 3.707
7 1.895 2.365 2.998 3.499 8 1.860 2.306 2.896 3.355
9 1.833 2.262 2.821 3.250 10 1.812 2.228 2.764 3.169
11 1.796 2.201 2.718 3.106 12 1.782 2.179 2.681 3.055
13 1.771 2.160 2.650 3.012 14 1.761 2.145 2.624 2.977
15 1.753 2.131 2.602 2.947 16 1.746 2.120 2.583 2.921
17 1.740 2.110 2.567 2.898 18 1.734 2.101 2.552 2.878
19 1.729 2.093 2.539 2.861 20 1.725 2.086 2.528 2.845
22 1.717 2.074 2.508 2.819 24 1.711 2.064 2.492 2.797
26 1.706 2.056 2.479 2.779 28 1.701 2.048 2.467 2.763
30 1.697 2.042 2.457 2.750 120 1.658 1.980 2.358 2.617

5.2.2. Estimation of Confidence Interval of Variance

Chi-Square (  2 ) distribution is used in the estimation of confidence interval of the variance. The probability
of  2 is greater than a given value of  0 is tabulated according to ( N  1) degree of freedom for various
2

confidence levels (Table 5.2). Confidence interval of the variance of the population (  2 ) is found as:

 NS 2 x NS 2 x 
Pc  Pb1   2  b2   P 2  2  2  (5.5a)
   / 2  1 / 2 
Confidence interval of the standard deviation of the population (  ) is found as follows:

 NS 2 x   2 
Pc  Pb1    b2   P       NS x  (5.5b)
   / 2    21 / 2 
2
  

In other words, the lower and upper limits of confidence interval of the standard deviation of the population
(  ) are:

NS 2 x NS 2 x
Lower limit: b1  , upper limit: b2  (5.5c)
 2 / 2  21 / 2
 41

TABLE 5.2. 2 DISTRIBUTION (df =  :DEGREE OF FREEDOM)

df =  = 0.99 0.975 0.95 0.90 0.10 0.05 0.025 0.01

 PC = 0.01 0.025 0.05 0.10 0.90 0.95 0.975 0.99
1 - - 0.004 0.0158 2.706 3.841 5.024 6.635
2 0.020 0.051 0.103 0.211 4.605 5.991 7.378 9.210
3 0.115 0.216 0.352 0.584 6.251 7.815 9.348 11.345
4 0.297 0.484 0.711 1.064 7.779 9.488 11.143 13.277
5 0.554 0.831 1.145 1.610 9.236 11.070 12.833 15.086
6 0.872 1.237 1.635 2.204 10.645 12.592 14.449 16.812
7 1.239 1.690 2.167 2.833 12.017 14.067 16.013 18.475
8 1.646 2.180 2.733 3.490 13.362 15.507 17.535 20.090
9 2.088 2.700 3.325 4.168 14.684 16.919 19.023 21.666
10 2.558 3.247 3.490 4.865 15.897 18.307 20.483 23.209
11 3.053 3.816 4.575 5.578 17.275 19.675 21.920 24.725
12 3.571 4.404 5.226 6.304 18.549 21.026 22.337 26.217
13 4.107 5.009 5.892 7.042 19.812 22.362 24.736 27.688
14 4.660 5.629 6.571 7.790 21.064 23.685 26.119 29.141
15 5.229 6.262 7.261 8.547 22.307 24.996 27.488 30.578
16 5.812 6.908 7.962 9.312 23.542 26.296 28.845 32.000
17 6.408 7.564 8.672 10.085 24.769 27.587 30.191 33.409
18 7.015 8.231 9.390 10.865 25.989 28.869 31.526 34.805
19 7.633 8.907 10.117 11.651 27.204 30.144 32.852 36.191
20 8.260 9.591 10.851 12.443 28.412 31.410 34.170 37.566
21 8.897 10.283 11.591 13.240 29.615 32.671 35.479 38.932
22 9.542 10.982 12.338 14.041 30.813 33.924 36.781 40.289
23 10.196 11.689 13.091 14.848 32.007 35.172 38.075 41.638
24 10.856 12.401 13.848 15.659 33.196 36.415 39.364 42.980
25 11.524 13.120 14.611 16.473 34.382 37.652 40.646 44.314
26 12.198 13.844 15.379 17.292 35.563 38.885 41.923 45.642
27 12.879 14.573 16.151 18.114 36.741 40.113 43.195 46.963
28 13.565 15.308 16.928 18.939 37.916 41.337 44.461 48.278
29 14.256 16.047 17.708 19.768 39.087 42.557 45.722 49.588
30 14.953 16.791 18.493 20.599 40.256 43.773 46.979 50.892
40 22.164 24.433 26.509 29.051 51.805 55.758 59.342 63.691
50 29.707 32.357 34.764 37.689 63.167 67.505 71.420 76.154
60 37.485 40.482 43.188 46.459 74.397 79.082 83.298 88.379
80 53.540 57.153 60.391 64.278 96.578 101.879 106.629 112.329
100 70.065 74.222 77.929 82.358 118.498 124.342 129.561 135.807
 42

5.3. EXAMPLES

Example 5.1: The mean and standard deviation values of compressive strength of 50 samples obtained from
a concrete mass are 240 kg/cm2 and 65 kg/cm2, respectively.
a. Determine the interval which contains 95 percent of mean compressive strength of concrete,
b. Calculate the probability that the mean compressive strength is greater than 260 kg/cm2.
Solution:

Example 5.2: Experiments were performed for 21 steel samples. It is found that mean and standard
deviation of yielding values in the experiments are 8490 kg and 100 kg, respectively. Calculate the
confidence intervals of population mean for a. 90 % and b. 98 % confidence levels.

Solution:
 43

Example 5.3: a. 20 and b. 10 samples are taken from concrete, produced a batch plant. Both groups have a
100 kg/cm2 standard deviation. Calculate the 95% confidence interval of population standard deviation for
both groups.
Solution:
 44

Example 5.4. The mean and standard deviation values of 20 samples taken from a concrete mass are

280 kg/cm2 and 15 kg/cm2, respectively. Calculate the confidence interval of population of a. mean and

b. standard deviation for 5 % significance level.

Solution:
 45

CHAPTER 6

STATISTICAL HYPOTHESIS

6.1. INTRODUCTION

In the previous chapter we illustrated how to construct a confidence interval estimate of a parameter from
sample data. However, many problems in engineering require that we decide whether to accept or reject a
statement about some parameter. The statement is called a hypothesis, and the decision-making procedure
about the hypothesis is called hypothesis testing. This is one of the most useful aspects of statistical
inference, since many types of decision-making problems, tests, or experiments in the engineering world can
be formulated as hypothesis-testing problems. Statistical hypothesis testing and confidence interval
estimation of parameters are the fundamental methods used at the data analysis stage of a comparative
experiment, in which the engineer is interested, for example, in comparing the mean of a population to a
specified value. These simple comparative experiments are frequently encountered in practice and provide a
good foundation for the more complex experimental design problems. Since we use probability distributions
to represent populations, a statistical hypothesis may also be thought of as a statement about the probability
distribution of a random variable. The hypothesis will usually involve one or more parameters of this
distribution. Hypotheses are always statements about the population or distribution under study, not
statements about the sample.
A procedure leading to a decision about a particular hypothesis is called a test of a hypothesis. Hypothesis
testing procedures rely on using the information in a random sample from the population of interest. If this
information is consistent with the hypothesis, we will conclude that the hypothesis is true; however, if this
information is inconsistent with the hypothesis, we will conclude that the hypothesis is false. We emphasize
that the truth or falsity of a particular hypothesis can never be known with certainty, unless we can examine
the entire population. This is usually impossible in most practical situations. Therefore, a hypothesis-testing
procedure should be developed with the probability of reaching a wrong conclusion in mind.
One of the main goals of the statistic is to perform correct estimations about population by using inadequate
data obtained from small samples. During this process, some assumes, called statistical hypothesis are done
and after some tests, their validity is accepted or rejected. Scientists collect data in order to learn about the
processes and systems those data represent. Often they have prior ideas, called hypotheses, of how the
systems behave. One of the primary purposes of collecting data is to test whether those hypotheses can be
substantiated, with evidence provided by the data.
Statistical tests are the most quantitative ways to determine whether hypotheses can be substantiated, or
whether they must be modified or rejected outright. The acceptance of a hypothesis does not imply that it is
certainly true and its rejection does not mean that it is certainly false. The determination of a hypothesis and
evaluating it as true or false can be expressed by some probabilities (90, 93, 99 percent, etc).
 46

One important use of hypothesis tests is to evaluate and compare groups of data. For example, water quality
has been compared between two or more aquifers, and some statements made as to which are different.
Rather than using hypothesis tests, the results are sometimes expressed as the author's educated opinions "it
is clear that development has increased well yield." Hypothesis tests have at least two advantages over
educated opinion:
i. They insure that every analyst of a data set using the same methods will arrive at the same result.
Computations can be checked on and agreed to by others,
ii. They present a measure of the strength of the evidence (the p-value).

6.2. STRUCTURE OF HYPOTHESIS TESTS

6.2.1. Choose an Appropriate Test

Test procedures are selected based on the data characteristics and study objectives. The second criterion is
the objective of the test. Hypothesis tests are available to detect differences between central values of two
groups, three or more groups, between spreads of data groups, and for covariance between two or more
variables, among others. The third selection criterion is the choice between parametric or nonparametric
tests. This should be based on the expected distribution of the data involved. If similar data in the past were
normally distributed, a parametric procedure would usually be selected. If data were expected to be non-
normal, or not enough is known to assume any specific distribution, nonparametric tests would be preferred.

6.2.2. Establish the Null and Alternative Hypotheses

The null and alternate hypotheses should be established prior to collecting data. These hypotheses are a
concise summary of the study objectives, and will keep those objectives in focus during data collection.
The null hypothesis (H0) is what is assumed to be true about the system under study prior to data collection,
until indicated otherwise. It usually states the null situation – no difference between groups, no relation
between variables. One may suspect, hope, or root for either the null or alternate hypothesis, depending on
one's vantage point. But the null hypothesis is what is assumed true until the data indicate that it is likely to
be false. For example, an engineer may test the hypothesis that wells upgradient and downgradient of a
hazardous waste site have the same concentrations of some contaminant. They may hope that downgradient
concentrations are higher (the company gets a new remediation project), or that they are the same (the
company did the original site design). In either case, the null hypothesis assumed to be true is the same:
concentrations are similar in both groups of wells.
The alternate hypothesis (H1) is the situation anticipated to be true if the evidence (the data) show that the
null hypothesis is unlikely. It is in some cases just the negation of H0, such as "the 100-year flood is not
equal to the design value." H1 may also be more specific than just the negation of H0 "the 100-year flood is
greater than the design value".
 47

6.2.3. Decide on an Acceptable Error Rate α

The α-value, or significance level, is the probability of incorrectly rejecting the null hypothesis H0 when it is
in fact true, called a "Type I error"). Table 6.1. shows that this is one of four possible outcomes of a
hypothesis test. The significance level is the risk of a Type I error deemed acceptable by the decision maker.
It is a " management tool" dependent not on the data, but on the objectives of the study. Statistical tradition

uses a default of 5% (0.05) for α, but there is no reason why other values should not be used. Before testing
the hypothesis, a significance level ( ) is chosen. This level represents the probability of type I error. If the
probability of type I error decreases, the probability of type II error increases, Therefore,  value can not be
chosen as small as desired.
Table 6.1. Four Possible Results of Hypothesis Testing.

Decision Unknown True Situation

H0 is True H0 is False

Accept H0 Correct Decision Type II Error

(Reject H1) Probability = 1-  Probability = 
Reject H0 Type I Error Correct Decision
(Accept H1) Probability =  Probability = 1- 

6.2.4. Make the Decision to Reject H0 or Not

When the p-value is less than the decision criteria (the α level), H0 is rejected. When the p value is greater
than α, H0 is not rejected. The null hypothesis is never accepted, or proven to be true. It is assumed to be true
until proven otherwise, and is not rejected when there is insufficient evidence to do so.

6.3. HYPOTHESIS TESTS FOR PARAMETERS

Hypothesis tests may be performed for values of parameters of population. The tests for the mean and the
standard deviation, the most widely used parameters in practice, are as follows:

6.3.1. Test of Mean

In order to test whether the means of two different populations are the same or not (H0: 1 = 2), samples are
taken from the populations with N1 and N 2 quantities, their means x1 and x 2 and standard deviations S1 and
S 2 are calculated. The standard normal value (z ) is found from normal distribution corresponding
to 0.5   / 2 ; the following values are calculated:

( N 1 S 2  N 2 S1 )
2 2

Sx  , x  x1  x2 (6.1)
N1 N 2

If x  zS x , then the H 0 is accepted, if x  zS x then it is rejected.

 48

6.3.2. Test of Standard Deviation

In order to test whether the standard deviations of two different populations are the same or not, (H0: 1 =2)
F (Fisher) test is used. The computed F is calculated as follows:
2 2
S1 S2
If S1  S 2  Fc  2
, if S 2  S1  Fc  2
(6.2)
S2 S1

If the computed Fc is less than its tabulated value (Ft) then H0 hypothesis is accepted, and vice versa. The
tabulated F is found from F Table (Table 6.1), with degrees of freedom (N1-1) and (N2-1) on numerator and
denominator, respectively. Here, N1 and N2 are the numbers of data of sample which have greater and
smaller variance, respectively.

TABLE 6.1. F DISTRIBUTION (   0.01)

n , m 
5 6 7 8 9 10 15 20 30 40 50 100
5 11.0 10.7 10.5 10.3 10.2 10.1 9.72 9.55 9.38 9.29 9.24 9.13
6 8.75 8.47 8.26 8.10 7.98 7.87 7.56 7.40 7.23 7.14 7.09 6.99
7 7.46 7.19 6.99 6.84 6.72 6.62 6.31 6.16 5.99 5.91 5.86 5.75
8 6.63 6.37 6.18 6.03 5.91 5.81 5.52 5.36 5.20 5.12 5.07 4.96
9 6.06 5.80 5.61 5.47 5.35 5.26 4.96 4.81 4.65 4.57 4.52 4.42
10 5.64 5.39 5.20 5.05 4.94 4.85 4.56 4.41 4.25 4.17 4.12 4.01
12 5.06 4.82 4.64 4.50 4.39 4.30 4.01 3.86 3.70 3.62 3.57 3.47
14 4.70 4.46 4.28 4.14 4.03 3.94 3.66 3.51 3.35 3.27 3.22 3.11
16 4.44 4.20 4.03 3.89 3.78 3.69 3.41 3.26 3.10 3.02 2.97 2.86
18 4.25 4.01 3.84 3.71 3.60 3.51 3.23 3.08 2.92 2.84 2.78 2.68
20 4.10 3.87 3.70 3.56 3.46 3.37 3.09 2.94 2.78 2.69 2.64 2.54
25 3.86 3.63 3.46 3.33 3.22 3.13 2.86 2.70 2.54 2.46 2.40 2.29
30 3.70 3.47 3.30 3.17 3.07 2.98 2.70 2.55 2.39 2.30 2.25 2.13
40 3.51 3.29 3.12 2.99 2.89 2.80 2.52 2.37 2.20 2.11 2.06 1.94
60 3.34 3.12 2.95 2.82 2.72 2.63 2.35 2.20 2.03 1.94 1.88 1.75
80 3.26 3.04 2.87 2.74 2.64 2.55 2.27 2.12 1.94 1.85 1.79 1.66
100 3.21 2.99 2.82 2.69 2.59 2.50 2.20 2.07 1.89 1.80 1.73 1.60
 49

6.4. HYPOTHESIS TESTS FOR PROBABILITY DISTRIBUTIONS

In order to test the frequency distribution of a sample whether or not fit a theoretical distribution (for
example normal distribution), test of statistical hypothesis is the most reliable method.

6.4.1. Chi Square Test

If, the observed sample with N components is classified into m classes; each of the class has Oi components
and the corresponding theoretical distribution has ei component; then 2 is calculated as follows:
m
 (Oi  ei ) 2 
 2
c    (6.3)
i 1  ei 

If  2 c   2 t then the sample fits the theoretical distribution, and vice versa. If  2 t  0 , exact fitting exist
between the sample and the related distribution. 2t is read for  . In reading the tabulated  2 values, the
degree of freedom is m-2 for Poisson and m-3 for other distributions.

6.4.2. Probability Plot Correlation Coefficient Test

In this test, the correlation coefficient (r) between the theoretical values of a distribution (xt) and observed
values (xo) is calculated. If the calculated value is greater than or equal to the critical value given in Table
6.2, then it is assumed that the observed data fit the related distribution. Correlation coefficient is found as:

r
N  xy   x  y

 x  x ( y  y )   xy  Nxy
N  x  
(6.4)
  x  N  y   y 
0.5
2 2 2 2 NS x S y NS x S y

The values are ranked from smaller to greater, the probability of non exceedance of each data is calculated
as following, where, i is the rank number.
i  0.40
F ( x  xi )  1  p  (6.5)
N  0.20
Table 6.2. Critical r Values
N G, P AND LP III N, LN
α=0.01 α=0.05 α=0.01 α=0.05
10 0.8630 0.9084 0.879 0.918
20 0.9060 0.9390 0.926 0.951
30 0.9191 0.9526 0.947 0.964
40 0.9286 0.9594 0.959 0.972
50 0.9389 0.9646 0.966 0.977
60 0.9467 0.9685 0.971 0.980
70 0.9506 0.9720 0.975 0.983
80 0.9525 0.9747 0.978 0.985
90 0.9554 0.9764 0.980 0.986
100 0.9596 0.9779 0.982 0.987
 50

6.5. EXAMPLES

Example 6.1. Lighting is applied to decrease traffic accidents in a junction. Accident numbers are given
below before lighting for 12 months (x1) and after lighting for 8 months (x2). Decide whether or not the
lighting has affected a. the mean and b. standard deviation of traffic accident for α = 0.01.

x1 5 8 11 7 6 9 6 8 6 7 7 10
x2 3 6 7 5 11 8 7 2 - - - -
Solution:
 51

Example 6.2: The number (N), mean ( x ) and standard deviations (Sx) of precipitation height
values before and after dam construction are given as follows. Determine whether or not dam construction
has changed a. the mean and b. the standard deviation of precipitation heights for α=0.01.

N x (cm) Sx (cm)

Before the dam 20 100 23

After the dam 25 116 10

Solution:
a.

b.
 52

Example 6.3: Numbers of occurring (x) and observation (O) of a flood, of which probability is 5 percent,
are given as follows, for a 40 yearly observation period. Determine whether the data fit Binom and Poisson
Distributions for 5% significance level.
Solution:
 53

Example 6.4: The 80 yearly annual total precipitation height values of a region fit Normal Distribution with
a mean of 70 cm and a standard deviation of 10 cm. It is aimed to test whether or not the data fit Normal
Distribution. There are 50 data between (60-75) cm, calculate the chi-square (χ2) value.
Solution:
 54

Example 6.5: The distribution of marks taken by 80 students in an exam is given as follows. The mean is 50
and the standard deviation is 16. Determine whether or not the marks fit normal distribution for both 1% and
5% significance levels.

GROUP 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100
NO OF 1 5 8 10 14 15 11 8 6 2
STUD. (Oi)

Solution:
x  x x  50
Number of data N = 80, number of groups m = 10, x  50 , S x  16 , N = 80, z   ,
Sx 16
For example, for the first group (0-10), the values are 0 and 10 and the corresponding z values are calculated
as follows:

If the data fits Normal Distribution, then, the probability of the data is between (0 - 10) is calculated as was

given in Normal Distribution:

0  50 10  50
x1  0  z1   3.13 , x 2  10  z 2   2.50 .
16 16

P(0  x  10)  P(3.13  z  2.50)  0.4991 0.4938  0.0053

Expected value ei = N*pi = 80*0.0053 = 0.424, observed value (no of student) Oi = 1

m
 (Oi  ei ) 2  1  0.424) 2 
 2
c     0.782 is obtained.
i 1  ei  0.424

For the second group (10-20), values are 10 and 20 and the corresponding z values are calculated as follows:

If the data fits Normal Distribution, then, the probability of the data is between (10 - 20) is calculated as was
given in Normal Distribution:
10  50 20  50
x1  10  z1   2.50 , x 2  20  z 2   1.88
16 16
P(10  x  20)  P(2.50  z  1.88)  0.4938  0.4699  0.0239
Expected value ei = N*pi = 80*0.0239 = 1.912, observed value (no of student) Oi = 5

m
 (Oi  ei ) 2  5  1.9122
 2
c     4.987 is obtained.
i 1  ei  1.912

Similar calculations are made for all of the groups and the results are presented in the table:
 55

GROUP 0-10 10-20 20-30 30-40 40-50

NO OF ST. (Oi) 1 5 8 10 14
x1 0 10 20 30 40

x2 10 20 30 40 50

z1 -3.13 -2.50 -1.88 -1.25 -0.63

z2 -2.50 -1.88 -1.25 -0.63 0.00

Pi 0.0053 0.0239 0.0755 0.1587 0.2357

ei=80*Pi 0.424 1.912 6.040 12.696 18.856

m
 (Oi  ei ) 2  0.782 4.987 0.636 0.572 1.251
 2c    
i 1  ei 

GROUP 50-60 60-70 70-80 80-90 90-100

NO OF ST. (Oi) 15 11 8 6 2
x1 50 60 70 80 90

x2 60 70 80 90 100

z1 0.00 0.63 1.25 1.88 2.50

z2 0.63 1.25 1.88 2.50 3.13

Pi 0.2357 0.1587 0.0755 0.0239 0.0053

ei = 80*Pi 18.856 12.696 6.040 1.912 0.424

m
 (Oi  ei ) 2  0.789 0.227 0.636 8.740 5.858
 2
c   
i 1  ei 

m
 (Oi  ei ) 2 
 2c      0.782  4.987  ...  8.740  5.858  24.478 is calculated.
i 1  ei 
If  2 c   2 t then the sample fits the theoretical distribution, and vice versa. 2t is read for 1-  . In reading

the tabulated  2 values, the degree of freedom is df = m-3 for Normal distribution. df = m – 3 = 10 – 3 =7

for 1% significance level (  = 0.01), 1 -  = 0.99, From Table 5.2   2 t  18.475 is read.
Conclusion: Since  2 c  24.878   2 t  18.475  The data do not fit Normal Distribution for 5% significance level.

for 5% significance level (  = 0.05), 1 -  = 0.95, From Table 5.2   2 t  14.067 is read.
Conclusion: Since  2 c  24.878   2 t  14.0675  The data do not fit Normal Distribution for 1% significance

level.
 56

Example 6.6: Classified values of 50 yearly mean annual discharges of a stream (m3/s) are given as
follows. The mean is 24.0 m3/s and the standard deviation is 3.80 m3/s.
a. Determine the 90 % confidence interval of mean of discharges for the stream population.

b. Suppose (assume) that the data fit normal distribution. If 100 yearly observations were made, estimate the
number of years which have greater than 20 m3/s discharge,

Solution:

c. Determine whether or not the data fit normal distribution for 1 % significance level.

Solution:
 57

Example 6.7: The annual maximum discharge values of a gauging station are given below. Test the data
whether or not fit Gumbel and Log Normal distributions by chi square (classifying into 10 groups) and
probability plot correlation coefficient tests for  =0.01 and  =0.05.
22.4 25.3 26.5 17.3 24.0 48.4 27.7 33.5 41.3 24.5 38.4 28.4 13.5 16.4 22.2 23.5 22.8 21.4 29.0 34.4
21.5 42.4 47.0 47.3 29.3 44.6 38.0 33.3 53.1 27.0 23.6 34.4 29.0 42.3 51.5 36.4 24.5 47.0 34.5 27.6
Solution:
a. BASIC CALCULATIONS:

The ranked values of x from smaller to greater for the original values (for Gumbel Distribution):
13.5 16.4 17.3 21.4 21.5 22.2 22.4 22.8 23.5 23.6 24.0 24.5 24.5 25.3 26.5 27.0 27.6 27.7 28.4 29.0
29.0 29.3 33.3 33.5 34.4 34.4 34.5 36.4 38.0 38.4 41.3 42.3 42.4 44.6 47.0 47.0 47.3 48.4 51.5 53.1
Number of data: N = 40, the mean x  31.88 , standard deviation S x  10.313 are calculated.
The ranked values of x from smaller to greater for the logarithmic values (y = lnx) (for Normal
Distribution):
2.60 2.80 2.85 3.06 3.07 3.10 3.11 3.13 3.16 3.16 3.18 3.20 3.20 3.23 3.28 3.30 3.32 3.32 3.35 3.37
3.37 3.38 3.51 3.51 3.54 3.54 3.54 3.59 3.64 3.65 3.72 3.74 3.75 3.80 3.85 3.85 3.86 3.88 3.94 3.97
Number of data: N = 40, the mean x  3.4105, standard deviation S x  0.3316are calculated.

b. CHI SQUARE TESTS:

b1: GUMBEL DISTRIBUTION:

x  31.88 , S x  10.313 , N = 40, From Gumbel Distribution Table 4.2  y N  0.544, S N 1.141,

 y  1.141  0.544 
  0.1106xi  26.963
SN
Eq. (4.19)  yi   xi  x  S x N    xi  31.88  10.313
Sx  S N  10.313  1.141 

 e 0.1106( xi 26.963)
Eq. (4.18a)  1  F ( xi )  P( x  xi )  e
e y
 e
Number of groups m = 10, the range is xmax – xmin = 53.1 – 13.5 = 39.6,

The group interval = The range/number of groups = 39.6/10 = 3.96,

The first group: (13.5) and (13.5 + 3.96 = 17.46)  (13.5 - 17.46),
The observed data between (13.5 and 17.46) are (13.5, 16.4 and 17.3) and number of data is Oi = 3
0.1106(13.526.963) 0.1106(17.4626.963)
P(x  13.5)  e e = 0.0119 ,
P(x  17.46) 
e e = 0.0572

P(13.5<x<17.46) = 0.0572 – 0.0119 = 0.0453, expected value ei = NPi = 40*0.0453 = 1.812


2

Oi  ei 2  3  1.8122  0.779
c
ei 1.812
 58

The second group (17.46) and (17.46 + 3.96 = 21.42)  (17.46 – 21.42),

The observed data between (17.46 and 21.42) is (21.4) and number of data is Oi = 1,

0.1106( 21.4226.963)
P(x  17.46)  0.1106(17.4626.963)
e e = 0.0572,
P(x  21.42)  e e = 0.1579,

P(17.46<x<21.42) = 0.1579 – 0.0572 = 0.1007, expected value ei = 40*0.1107 = 4.028


2

Oi  ei 2  1  4.0282  2.276
c
ei 4.028

Similar calculations are made for all of the groups and presented in the following table:

GROUP

Oi

x1

x2

P2

P1

 P = P2 – P1

ei =  P*40


2

Oi  ei 
2

c
ei
 59

b2: LOG NORMAL DISTRIBUTION:

Number of data: N = 40, the mean x  3.4105, standard deviation S x  0.3316

x  x x  3.4105
Eq. (4.11)  z   ,
Sx 0.3316
Number of groups m = 10, range is xmax – xmin = 3.97 – 2.60 = 1.37, the group interval = 1.37/10 = 0.137,
The first group: (2.60) and (2.60 + 0.137 = 2.737)  (2.60 – 2.737),
The observed data between (2.60 and 2.737) is (2.60) and number of data is Oi = 1
2.60  3.4105 2.737  3.4105
For the first group:  z1   2.44 , z 2   2.03
0.3316 0.3316
P(2.60  x  2.737)  P(2.44  z  2.03)  0.4927  0.4788  0.0139 ,

The expected value ei = 40*0.0139 = 0.556, 

2

Oi  ei 2  1  0.5562  0.355
c
ei 0.556

The second group: (2.737) and (2.737 + 0.137 = 2.874)  (2.737 – 2.874),
The observed data between (2.737 and 2.874) are (2.80 and 2.85) and number of data is Oi = 2
2.737  3.4105 2.874  3.4105
For the second group: z1   2.03 z 2   1.62
0.3316 0.3316
P(2.737  x  2.874)  P(2.03  z  1.62)  0.4788  0.4474  0.0314 ,

The expected value ei = 40*0.0314 =1.256, 

2

Oi  ei 2  2  1.2562  0.441.
c
ei 1.256

Similar calculations are made for all of the groups and presented in the following table:

GROUP

Oi

z1

z2

P1

P2

 P = P2 – P1

ei =  P*40


2

Oi  ei 2
c
ei
 60

c. PROBABILITY PLOT CORRELATION COEFFICIENT TESTS:

The correlation coefficient (r) between the theoretical values of a distribution (x t) and observed values (xo) is
calculated. If the calculated value is greater than or equal to the critical value given in Table 5.2, then it is
assumed that the observed data fit the related distribution. The correlation coefficient between x and y is
calculated by Eq. (5.4):

r
N  xy   x  y

 x  x ( y  y )   xy  Nxy
N  x 2 2

  x  N  y 2   y 
2

0.5
NS x S y NS x S y

The values are ranked from smaller to greater, the probability of non exceedance of each data is calculated
by Eq. (5.5) where, i is the rank number.
i  0.40
F ( x  xi )  1  p 
N  0.20
When the data are ranked from minimum value to maximum value, N = 40, for the minimum value i = 1,
for the second value i = 2, ... for the maximum value i = 40.
1  0.40
For the minimum value the probability of value less than minimum value is F ( x  xi )   0.0149 ,
40  0.20
2  0.40
For the second value the probability of value less than second value is F ( x  xi )   0.0398,
40  0.20
For the maximum value the probability of value less than maximum value is
40  0.40
F ( x  xi )   0.9851.
40  0.20
c1: GUMBEL DISTRIBUTION:
N = 40  y N  0.544, S N 1.141, x  31.88 , S x  10.313

Eq. (3.14a)  P( x  xi )  e  e  y   ln ln(( P( x  xi )) ,

y

For example for the minimum value y   ln ln((0.0149)  1.437 ,

For the second value y   ln ln((0.0398)  1.171,
For the maximum value y   ln ln((0.9851)  4.199 .
Sx y 10.313 0.544
Eq. (3.15a) xi  yi  x  Sx N  yi  31.88  10.313  xi  9.039yi  26.963
SN S N 1.141 1.141

For the minimum value, y1  1.437  x1  9.039 * (1.437)  26.963  13.99

This is the expected value for the minimum discharge, e1 = 13.99, the observed minimum discharge is O1 =13.5

For the second value y2  1.171 x2  9.039* (1.171)  26.963  16.39

 61

This is the expected value for the second discharge, e2 = 16.39, the observed minimum discharge is O2 = 16.4

For the maximum value, y40  4.199  x1  9.039* (4.199)  26.963  64.9

This is the expected value for the maximum discharge, e40= 64.9, the observed minimum discharge is O40 =53.1

Similar calculations are made for all of the groups and presented in the following table:

i 1 2 3 4 5 6 7 8 9 10 11

Pi

Oi

yi

ei

i 12 13 14 15 16 17 18 19 20 21 22

Pi 0.2886 0.3134 0.3383 0.3632 0.3881 0.4129 0.4378 0.4627 0.4876 0.5124 0.5373

Oi 24.5 24.5 25.3 26.5 27.0 27.6 27.7 28.4 29.0 29.0 29.3

yi -0.217 -0.149 -0.080 -0.013 0.055 0.123 0.191 0.260 0.331 0.402 0.476

ei 25.00 25.62 26.24 26.85 27.46 28.07 28.69 29.32 29.96 30.60 31.27

i 23 24 25 26 27 28 29 30 31 32 33

Pi 0.5622 0.5871 0.6119 0.6368 0.6617 0.6866 0.7114 0.7363 0.7612 0.7861 0.8109

Oi 33.3 33.5 34.4 34.4 34.5 36.4 38.0 38.4 41.3 42.3 42.4

yi 0.552 0.631 0.711 0.796 0.884 0.978 1.077 1.184 1.299 1.424 1.563

ei 31.95 32.66 33.39 34.15 34.96 35.80 36.70 37.66 38.70 39.84 41.09

i 34 35 36 37 38 39 40

Pi 0.8358 0.8607 0.8856 0.9104 0.9353 0.9602 0.9851

Oi 44.6 47.0 47.0 47.3 48.4 51.5 53.1

yi 1.718 1.897 2.108 2.366 2.705 3.204 4.199

ei 42.49 44.11 46.02 48.35 51.41 55.92 64.92

 62

c2: LOG NORMAL DISTRIBUTION:

Number of data: N = 40, the mean x  3.4105, standard deviation S x  0.3316

x  x x  3.4105
Eq. (3.7)  z    x  0.3316z  3.4105 ,
Sx 0.3316

For the minimum value, P = 0.0149 < 0.5  z  0 , from table 1, P = 0.5 – 0.0149 = 0.4851  z  2.17

 x  0.3316(2.17)  3.4105  2.69

This is the expected value for the minimum discharge, e1 = 2.69, the observed minimum value is O1 =2.60,

For the second value, P = 0.0398 < 0.5  z  0 , from table 1, P = 0.5 – 0.0398 = 0.4602  z  1.75

 x  0.3316(1.75)  3.4105  2.83

This is the expected value for the second discharge, e2 = 2.69, the observed second value is O2 =2.80,

For the maximum value, P = 0.9851 > 0.5  z  0 , from table 1, P = 0.9851 – 0.5 = 0.4851  z  2.17

 x  0.3316(2.17)  3.4105  4.13

This is the expected value for the maximum discharge, e40 = 4.13, the observed maximum value is O40 = 3.97,

Similar calculations are made for all of the groups and presented in the following table:
 63

i 1 2 3 4 5 6 7 8 9 10 11

Pi

Oi

zi

ei

i 12 13 14 15 16 17 18 19 20 21 22

Pi 0.2886 0.3134 0.3383 0.3632 0.3881 0.4129 0.4378 0.4627 0.4876 0.5124 0.5373

Oi 3.20 3.20 3.23 3.28 3.30 3.32 3.32 3.35 3.37 3.37 3.38

zi -0.56 -0.49 -0.42 -0.35 -0.30 -0.22 -0.16 -0.09 -0.03 0.03 0.09

ei 3.22 3.25 3.27 3.29 3.31 3.34 3.36 3.38 3.40 3.42 3.44

i 23 24 25 26 27 28 29 30 31 32 33

Pi 0.5622 0.5871 0.6119 0.6368 0.6617 0.6866 0.7114 0.7363 0.7612 0.7861 0.8109

Oi 3.51 3.51 3.54 3.54 3.54 3.59 3.64 3.65 3.72 3.74 3.75

zi 0.16 0.22 0.30 0.35 0.42 0.49 0.56 0.63 0.71 0.79 0.88

ei 3.46 3.48 3.50 3.52 3.54 3.56 3.58 3.61 3.63 3.66 3.68

i 34 35 36 37 38 39 40

Pi 0.8358 0.8607 0.8856 0.9104 0.9353 0.9602 0.9851

Oi 3.80 3.85 3.85 3.86 3.88 3.94 3.97

zi 0.98 1.09 1.20 1.34 1.52 1.75 2.17

ei 3.72 3.75 3.81 3.85 3.92 3.99 4.13

 64

CHAPTER 7

REGRESSION ANALYSIS

7.1. GENERAL INFORMATION

7.1.1. Introduction

Most of the variables in engineering are statistically dependent of each other, thus there is a relation between
these variables. For example, there is relation between strength of concrete and its cement ratio or between
flow and precipitation of a basin. One of these variables (for example precipitation) is defined as
independent variable and the other variable (flow) is called dependent variable of the independent variable.
The reason for such a relation is either one variable is affected by the other or both variables are affected by
other variable(s). As an example, the relation between precipitation and flow in a basin originates because
flow takes place due to effect (as a consequence) of precipitation. The relation between flows in neighboring
basins arises due to fact that the flows are affected by the precipitation of that region.
The relations have not a deterministic (functional) character; in other words, when one of the variables takes
a certain value, the other will not always take the same value. This value will change more or less in various
observations with the effect of other variables which have not been considered in the relation. Moreover, the
estimations depend on the very limited data of the sample, of which degree of representation to the
population is questionable. For example, when flow of one of the neighboring basins takes a certain value,
the flow of the other basin does not always take the same value. Nevertheless, the determination of the
existence and the form of a nonfunctional relationship between the variables has a great importance in
practice. Because, by using this relationship, it is possible to estimate a future value of a variable depending
on known value(s) of another (or more than one) variable(s). While this estimate will not be the exact future
value of the variable under consideration, it will be the best estimate closest to this value. The mathematical
expression showing a relation of the above mentioned type is called the regression equation. The aims of the
regression analysis are:
 To check whether there is a significant relation between the variables under consideration, and if
there is one, to obtain regression equation expressing this relation,
 To study and evaluate the reliability of the estimates to be made by using this equation.
The relation between flows of two neighboring basins recorded for same years is an example to the usage of
regression analysis in Civil Engineering. If the regression equation can be obtained, the missing data
unrecorded in the past in one of the basins can be estimated by using this equation. The estimated values are
the best ones, though they are not equal to the real values which would have been observed.
 65

7.1.2. Types of Regression

Depending on the number of independent variable(s) and the type of relation, there are three types of
regression equations:
a. Simple Linear Regression: There is one dependent and one independent variables and the relation between
these variables is linear. This is the simplest type of regression equations.
b. Multivariate Linear Regression: The number of independent variables is greater than one and the relation
is linear.
c. Nonlinear Regression: The relation between dependent variable and independent variable(s) is nonlinear
(for example parabolic, exponential etc). The regression is transformed into linear form by using appropriate
transformation techniques.

7.2. SIMPLE LINEAR REGRESSION

Simple linear regression equation is as follows:

y  a  bx (7.1)

Where, y is dependent and x is independent variables, a and b are regression coefficients. When x and y
values are plotted on a coordinate, if the data are scattered around a line, it is supposed that there is a linear
regression between x and y (Figure 7.1).
y
y  a  bx

* *
* b *
* * * 1
* * *
a *
a/b x

Figure 7.1 Regression Line

7.2.1. Correlation Coefficient

Correlation coefficient (r) shows the degree of the relation and the reliability of the regression equation
between x and y values. r values change between the interval of -1 and +1. As the reliability of the
regression equation increases, the absolute value of r approaches 1, which also explains that the relation and
dependence between x and y are strong. If r value is positive (negative) a positive (negative) relation there
exist between x and y; if it is equal to zero (r  0) , it is obvious that there is no relation. r  1 and r  1
imply that, there is exact positive and negative (the values of a variable increase while values of other
variable decrease) relations. When the observed x and y values are plotted, except the situation r  1 , a
scatter is observed. As the scattering increases, r approaches zero; in the case of haphazard scattering
correlation coefficient is nearly equal to zero (Figure 7.2).
 66

An estimation value for correlation coefficient of the population (  ) is calculated as follows:

r
N  xy   x  y

 x  x ( y  y )   xy  Nxy
N  x  
(7.2)
  x  N  y 2   y 
0.5
2 2 2 NS x S y NS x S y

Note that, in this equation, during the calculation of Sx and Sy values, even if the number of data is less than
30, S x    x  x  / N and S y    y  y  / N equations should be used and the value of (N-1) should
2 2 2 2

not be used instead of N.

y y y y
* * * * *
* ** * * * * *
* * * * * * *
* * * * * *
x x x x
a. r>0 b. r <0 c. r1 d. r-1
y y y y
* * * * * * * * * *
* ** * * * * * * * * * * **
* * * * * * * * * ** * * * * * * *
* * * * * ** * * * * ** *
x x x x
e. r0.75 f. r 0.5 g. r0.25 h. r0
Figure 7.2. Correlation Coefficient Values for Various Scattering Scenarios
Even if the correlation coefficient calculated from the sample is different from zero (r0), as a result of
sampling errors, it is possible that the correlation coefficient of the population is equal to zero (=0). In
order to check this, the calculated r values are compared to the critical values (r cr ) , which are tabulated
according to significance level () and degree of freedom (df = N-2) (Table 7.1). If r  rcr , then it is
concluded that the correlation coefficient is not equal to zero and that the regression equation is reliable.
Table 7.1. Critical Values of Correlatıon Coeffıcıent (df=N-2)
 
df 0.05 0.01 df 0.05 0.01
1 .999 .999 2 .950 .999
3 .878 .959 4 .811 .917
5 .754 .875 6 .707 .834
7 .666 .798 8 .632 .765
9 .602 .735 10 .576 .708
11 .553 .684 12 .532 .661
13 .514 .641 14 .497 .623
15 .482 .606 16 .468 .590
17 .456 .575 18 .444 .561
19 .433 .549 20 .423 .537
22 .404 .515 24 .388 .496
26 .374 .479 28 .361 .463
30 .349 .304 40 .304 .393
50 .273 .354 100 .195 .254
 67

7.2.2. Regression Equation

The regression coefficients of simple linear regression equation (a and b) are calculated as follows: The sum
of the squares of the vertical distance values (e) between the data and regression line should be minimum
(Figure 7.3). Therefore, the following function should be minimal:

  y  y     y  a  bx    y  a  bx
 2
e 2 2 2
(7.3)

The partial derivatives of a and b should be equal to zero, separately:

e
 2  y  a  bx 1  2  y  a  bx  0 (7.4a)
a
e
b

 2  y  a  bx x   2  xy  ax  bx 2  0  (7.4b)

y *

y i = a + bxi ,


ei y i  ei = yi  yi  ei = i

yi *

yi ei  y i  y i
x
Figure 7.3 Vertical Distances from Regression Equation
From these equations one can obtain:

  y  a  bx  0    y  Na  b x  0 (7.5a)

  xy  ax  bx   0
2
   xy  a x  b x 2  0 (7.5b)

The following equations are obtained from these equations:

 y  Na  b x (7.6a)

 xy  a x  b x 2
(7.6b)

These equations are called normal equations. By solving them, a and b are calculated as:

b
 x  x  y  y   N  xy   x y  r S y
, (7.7a)
 x  x  N  x   x 
2 2
2 S x

a  y  bx (7.7b)

Note that; in the last expression of (7.7a), since the standard deviation values are always positive, b values
may be positive (negative) if r values are positive (negative).
 68

NOTE: If (7.1) ( y  a  bx ) is summed from 1 to N, the following equation is obtained;

 y   a  b x  Na  b x (7.6a)

and if it is multiplied by x and ( xy  ax  bx 2 ) and is summed the following equation is obtained:

 xy   ax bx 2
 a  x  b x 2 (7.6b)

7.3. MULTIVARIATE LINEAR REGRESSION

The multivariate regression equation is as follows:

K
y  a0  a1 x1  a2 x2  ....  ak xk  a0   ai xi (7.8)
i 1

If the number of independent variable is K=2, then the equation is as follows:

y  a0  a1 x1  a 2 x 2 (7.9)

If this equation is summed from 1 to N, then is multiplied by x1 and x2 and summed from 1 to N, the
following normal equations are obtained:

y  a0  a1 x1  a 2 x 2   y  Na 0  a1  x1  a 2  x 2 (7.10a)

yx1  a0 x1  a1 x1  a 2 x1 x2   yx  a0  x1  a1  x1  a2  x1 x 2
2 2
1 (7.10b)

yx2  a0 x2  a1 x1 x2  a2 x2   yx  a0  x 2  a1  x1 x2  a2  x 2
2 2
2 (7.10c)

By solving these equations, the regression coefficients a 0 , a1 , a 2 are calculated and the regression equation is
obtained.

Multivariate correlation coefficient is calculated as:

0.5
 Se
2 

R  1 2  (7.11)
 S 
 y 

Se2 is the variance of the vertical distances from the regression equation (e) and is calculated as:

e  y  yc 
2 2

Se  
2 r
(7.12a)
N  K 1 N  K 1

Since the number of independent variable K=2, the equation is obtained as:

e  y  yc 
2 2

 
2 r
Se (7.12b)
N 3 N 3
 69

Here, e values are the differences between real and calculated y values (yr and yc); Sy2 is the variance of
real y values.

NOTE: If the number of independent variables is greater than 2, similar calculations are made. For example:
In the case of 3 independent variables (K=3), the regression equation is as follows:

y  a0  a1 x1  a 2 x 2  a3 x3 (7.13)

Sum the equation from 1to N:  y  Na 0  a1  x1  a 2  x 2  a3  x 3

x y  a x  a1  x1  a2  x1 x2  a3  x1 x3
2
Multiply by x1 and sum 1 0 1

x y  a0  x2  a1  x1 x2  a2  x2  a3  x2 x3
2
Multiply by x2 and sum 2

x y  a x  a1  x1 x3  a2  x2 x3  a3  x 3
2
Multiply by x3 and sum 3 0 3 (7.14)

By solving these equations, a0 , a1 , a 2 , a3 are found.

7.4. NONLINEAR REGRESSION

The linear regression model is the model most frequently used because of its simplicity; however, its usage
may be erroneous in the case of very small correlation coefficients and then nonlinear regression analysis
should be employed. There are numerous types of nonlinear regression. In the analysis of all of them, the
regression is transformed into linear form by using appropriate transformation techniques. The normal
equations are established and both regression coefficients (a0 , a1 ,..a K ) and correlation coefficient (r ) are
calculated. Some of the nonlinear regression types that are frequently encountered and their normal
equations are given in the following. In the analysis of other nonlinear regressions, similar principles are
employed:

7.4.1. Polynomial Function

The most common form of a polynomial is as follows:

y  a0  a1 x  a2 x 2  a3 x 3  ...  aK x K (7.15a)

If K=2, the polynomial is called parabola:

y  a0  a1 x  a2 x 2 (7.15b)

If the transformations of x1  x and x 2  x 2 are made, the equation is obtained y  a0  a1 x1  a 2 x 2 , as a

multivariate linear regression (Eq. 7.9). Normal equations are found by Eq. (7.10 a, b c) and regression
coefficients ( a 0 , a1 , a 2 ) are calculated. The multivariate correlation coefficient is calculated by (7.11). In the
cases of K is greater than 2, similar procedure is employed.
 70

7.4.2. Exponential Function

Exponential function is as follows:

y  ab x (7.16a)
This equation is made linear by using the following transformation:
log y  log a  x log b (7.16b)
Let Y  log y , A  log a and B  log b , the equation is transformed to a linear one:
Y  A  Bx (7.16c)

Correlation coefficient is calculated by Eq. (6.2). If the Eq. (6.16c) is summed and then multiplied by x and
summed, its normal equations are obtained as:

 Y  NA  B x ,  xY  A x  B x 2
(7.16d)

By using these equations, firstly A and B and then a  10 A and b  10 B regression coefficients are found
and inserted in Eq. (7.16a), thus, the regression equation is obtained.

7.4.3. Hyperbolic Function

Hyperbolic function is as follows:

y  ax b  log y  log a  b log x (7.17a)

If Y  log y , A  log a and X  log x are written, then the equation is linearized as:
Y  A  bX (7.17b)
Correlation coefficient is calculated by Eq. (7.2). If Eq. (7.17b) is summed and then multiplied by X and
summed, normal equations are obtained as follows:

Y  NA  b X ,  XY  A X  b X 2
(7.17c)

By using these equations, A (and a  10 A ) and b are found, by inserting these values in Eq. (7.17a), the
regression equation is obtained.

7.4.4. Geometric Function

Geometric function is as follows:

1 1
y   a  bx (7.18a)
a  bx y
If Y  1 / y is transformation is made, Eq. (7.18b) is obtained as a linear one:

Y  a  bx (7.18b)
Correlation coefficient is calculated by Eq. (7.2). If Eq. (7.18b) is summed and then multiplied by x and
summed, normal equations are obtained as follows:

Y  Na  b x ,  xY  a x  b x 2
(7.18c)

By using these equations, a and b are found, by inserting these values in Eq. (7.18a), the regression
equation is obtained.
 71

7.5. EXAMPLES

Example 7.1: There is simple linear regression between the following x and y values. Calculate the
correlation coefficient, test its reliability (whether correlation coefficient of the population is equal to zero)
for 0.01 and 0.05 significance levels, and if it is suitable, calculate the regression equation; by using this
equation estimate the y value for x = 5, and x value for y =14.
Solution: 
x 22 19 16 13 9 7 4 90
y 8 11 15 17 15 16 20 102

Eq. (7.2)  r 
N  xy   x y

 x  x ( y  y )   xy  Nxy
N  x 2 2

  x  N  y 2   y 
2

0.5
NS x S y NS x S y

Calculation of correlation coefficient: Only the last equation ( r 

 xy  Nxy ) will be used.
NS x S y

N  7, x  90 / 7  12.857, S x  6.081, y  102/ 7  14.571, S y  3.659

 72

Example 7.2: Linear regression is supposed between the 7 daily and 28 daily strength values (kg/cm 2) of a
concrete population. The measured 7 daily (x) and 28 daily (y) strength values of 10 samples are given
below.

a. Calculate correlation coefficient and test its reliability for α=0.01 and α=0.05 significance levels,
b. Obtain the regression equation and estimate 28 daily (y) strength values for x=280 kg/cm2 and
x= 300 kg/cm2 and also estimate x values for y= 210 kg/cm2 and 250 kg/cm2.

x 230 245 225 256 238 219 227 243 234 228
y 290 305 280 315 286 280 284 302 289 275

Data Solution: a.
y xy xx yy ( x  x) 2 ( y  y) 2 ( x  x).( y  y)
2 2
x x y
230 290 66 700 52 900 84 100 -4.5 -0.6 20.25 0.36 2.7
245 305 74 425 60 025 93 025 10.5 14.4 110.25 207.36 151.2
225 280 63 000 50 625 78 400 -9.5 -10.6 90.25 112.36 100.7
256 315 80 640 65 536 99 225 21.5 24.4 462.25 595.36 524.6
238 286 68 068 56 644 81 796 3.5 -4.6 12.25 21.16 -16.1
219 280 61 320 47 961 78 400 -15.5 -10.6 240.25 112.36 164.3
227 284 64 468 51 529 80 656 -7.5 -6.6 56.25 43.56 49.5
243 302 73 386 59 049 91 204 8.5 11.4 72.25 129.96 96.9
234 289 67 626 54 756 83 521 -0.5 -1.6 0.25 2.56 0.8
228 275 62 700 51 984 75 625 -6.5 -15.6 42.25 243.36 101.4

2345 2906 682633 551009 845952 0 0 1106.5 1468.4 1176.0

N = 10, x  234.5 , y  290.6 , Sx = 10.52, Sy = 12.12

 73

Eq. (7.2)

r
N  xy   x  y

 x  x ( y  y )   xy  Nxy 
N  x 2 2

  x  N  y 2   y  
2 0.5 NS x S y NS x S y

10 * 682633 2345* 2906 1176 682633 10 * 234.5 * 290.6

  
10 * 55109 2345 10 * 845952 2906 
2 2 0.5
10 *10.52 *12.52 10 *10.52 *12.52

 r  0.923, Absolute value r  0.923 , from Table 7.1, df = N – 2 = 10 – 2 = 8,

For 0.01 significance level (  =0.01)  rcr = 0.765, r  0.923 > rcr, therefore, the correlation coefficient of

the population is not equal to zero, the this value is reliable,

For 0.05 significance level (  =0.05)  rcr = 0.632, r  0.923 > rcr, therefore, the correlation coefficient of

the population is not equal to zero, the this value is reliable.

Calculation of regression coefficient and obtaining of regression equation:

Eq. (7.1): y  a  bx

Normal equations (6.6a and 6.6b):  y  Na  b x  2906  10a  2345b ,

 xy  a x  b x 2
 682633 2345a  551009b

By solving these equations one may obtain a = 41.327 and b = 1.063 are obtained.

Or by using Eq. (7.7a)

Sy 12.12
br  0.923  1.063 , a  y  bx  290.6  1.063* 234.5  41.327
Sx 10.52

Regression Equation: y  a  bx  41.327  1.063x ,

For x = 210  y  41.327  1.063* 210  264.6kg / cm ,

2

For x = 250  y  41.327  1.063* 250  307.1kg / cm

2

For y = 280  y  280  41.237  1.063x  x  224.5kg / cm and

2

For y = 300  y  300  41.237  1.063x  x  243.3kg / cm

2

values are obtained.

 74

Example 7.3: The equation among the following data is y  x1  2 x2  3x3  4 x4 . Calculate the correlation
coefficient and test its reliability for α=0.05 and α=0.01 significance levels.

x1 1 2 2 3 3 3 3
x2 2 2 3 3 3 4 4
GIVEN x3 3 4 3 4 5 4 5
DATA x4 4 4 5 5 4 3 6
y = yr -3 3 -3 0 6 10 2
Solution:
CALCU-
LATIONS
 75

Example 7.4: Multiple linear regression is supposed between the mean annual precipitation height values of
A gauge station (y) and B (x1) and C (x2) gauge stations ( y  a0  a1 x1  a 2 x 2 ). Observed values for 6 years
(cm) are given below.
a. Obtain the regression equation, estimate the mean annual precipitation value (y) in A gauge station for
x1=150 cm and x2= 80 cm.
b. Calculate the multiple correlation coefficient.

y 120.3 140.5 135.2 105.4 138.3 146.5

x1 183.5 210.4 201.3 163.2 214.0 260.3
x2 96.3 101.5 98.0 88.5 103.1 118.2
Solution: a.
x1 x2 y x1* x2 x1*y x2*y x12 x22
183.5 96.3 120.3 17671 22075 11585 33672 9274
210.4 101.5 140.5
201.3 98.0 135.2

260.3 118.2 146.5

1233 605.6 786.2 126028 163783 79998 258648 61613

 y  Na 0 a x 1 1
 a2  x 2  786.2 = 6a0 + 1233a1 + 605.6a2

 yx  a0  x1  a1  x1  a2  x1 x 2 
2
1 163783 = 1233a0 + 258648a1 + 126028a2

 yx  a0  x 2  a1  x1 x2  a2  x 2 
2
2 79998 = 606a0 + 126028a1 + 61613a2

By solving these equations, one can obtain a0 = 183.984 a1 = 1.465 a2 = -3.506

and regression equation is y  183.984  1.465x1  3.506x2
y  183.984  1.465 *150  3.506 * 80  123.254cm
b.
yr x1 x2 yc e e2
120.3 183.5 96.3 115.18 5.12 26.21
140.5 210.4 101.5
135.2 201.3 98.0
105.4
138.3
146.5
146.5 260.3 118.2 150.91 -4.41 19.45
 0.0 122.57

e  y  yc 
2 2
122.57
Sy2 = 195.2 , S e 2   r
  40.86 ,
N  K 1 N  K 1 6  2 1
0.5
 S  2
 40.86 
0.5

R  1  e 2   1    0.889
 S   195.2 
 y 
 76

Example 7.5: There is an equation of y  2  3x1  2 x2  x3 among the following data. Obtain the correlation
coefficient.

x1 1 2 3 4 5 6
x2 2 4 2 2 10 10
x3 2 3 1 2 2 1
y 6 5 -3 -7 10 9
Solution:
yc
e
e2

0.5
 16 , R  1 
16 
 e 2  32 , S y  38.889 ,  e 2  32 , S e 2 
2 32
  0.767
6  3 1  38.889 

a bx3
Example 7.6. There is a regression equation of y  e among the following data. Calculate the
correlation coefficient; test its reliability for 0.01 and 0.05 significance levels. Obtain the regression
equation, calculate y value for x = 0.7 and x value for y = 1.7

x -0.5 0.0 0.5 1.0 1.2 1.3

y 2.1 2.0 1.9 1.5 1.3 1.2
Solution:
y  e abx  ln y  a  bx3 , Y  ln y, X  x 3  Y  a  bX
3

X = x3 -0.125 0.000 0.125 1.000 1.728 2.197

Y = lny 0.742 0.693 0.642 0.405 0.262 0.182
X*Y -.0927 .000 .0802 .4055 .4534 .4006

 X  4.925,  Y  2.927,  XY  1.2469 , X  0.8208, S X  0.8946, Y  0.4878, S Y  0.2166

1.2469  6 * 0.8208* 0.4878

r  0.994 , r  0.994 , N = 6, df = N – 2 = 6 – 2 =4, from Table 7.1  for
6 * 0.8946 * 0.2166
 =0.05 rc = 0.811 < r  0.994  It is reliable,

 for  =0.01 rc = 0.917 < r  0.994  It is reliable

SY 0.2166
br  0.994  0.2407 , a  Y  bX  0.4878  (0.2407) * 0.8208  0.6853
SX 0.8946

y  e abx  e 0.68530.2407 x
3 3
The regression equation is: ,

x  0.7  y  e0.6853 0.2407*0.7  1.827 ,

3
For
0.6853 0.2407*0.7 3
for y  1.7  e  ln1.7  0.5306  0.6853  0.2407* x3  x  0.863
 77

Example 7.7: There is a regression equation of y  ln(a  bx 2 ) between the following x and y values.
Determine the regression equation and estimate y value for x=0.5. Calculate the correlation coefficient and
test its reliability for α=0.05 and α=0.01.

x 1.0 1.5 2.0 2.5 3.0

y 0.7 0.5 0.25 0.0 -0.3
Solution:
y  ln(a  bx2 )  e y  a  bx2 , if the transformations of Y = ey and X = x2 are made, the following
equation is obtained: e y  Y  a  bx 2  a  bX , Then, a simple linear equation between X = x2 and Y = ey is
carried out.

Example 7.8: There is

a. Linear, b. Equation of y  log(a  bx0.5 )
between the following variables. For both of the equations, calculate the correlation coefficients; obtain the
regression equations and estimate y values for x=2.0 and x values for y=0.5.

x 1.3 1.5 1.8 2.4 2.7

y 0.8 0.63 0.55 0.43 0.35
Solution:
 78

Example 7.9: Obtain the a. Exponential, b. Hyperbolic functions for the following x and y values. Calculate
correlation coefficients and determine which one more reliable is.
Solution:
a. Exponential function: Eq. (7.16a)  y  ab x
This equation is made linear by using the following transformation: log y  log a  x log b
Let Y  log y , A  log a and B  log b , the equation is transformed to linear: Y  A  Bx


GIVEN x 0.5 1.0 1.5 2.0 2.5 3.0 10.5
DATA y 2 8 18 30 70 100 228
SOLU- Y = log(y) 0.301 0.903 1.255 1.477 1.845 2.000 7.782
TION: x*Y 0.151 0.903 1.883 2.954 4.613 6.000 16.503

Number of data N = 6, basic calculations on x and Y yield:

x  10.5 / 6  1.75, S x  0.854 , Y  7.782 / 6  1.297 , SY  0.574 ,  x *Y  16.503
r
 xY  NxY 
16.503  6 *1.75 *1.297
 0.981
NS x SY 6 * 0.854 * 0.574
SY 0.574
B  log b  r  0.981  0.6592  b  10B  100.6592  4.562
Sx 0.854
A  log a  Y  Bx  1.297  0.6592*1.75  0.1434  a  10 A  100.1434  1.391
Regression equation: y  ab x  1.391* 4.562x

b. Hyperbolic Function: Eq. (7.17a)  y  ax b

This equation is made linear by using the following transformation: log y  log a  b log x
Let Y  log y , A  log a and X  log x , the equation is transformed to linear: Y  A  bX

GIVEN x 0.5 1.0 1.5 2.0 2.5 3.0 -
DATA y 2 8 18 30 70 100 -
SOLU- Y = log(y) 0.301 0.903 1.255 1.477 1.845 2.000 7.782
TION: X = log(x) -0.301 0.000 0.176 0.301 0.398 0.477 1.051
X*Y -0.0906 0.0000 0.2209 0.4446 0.7343 0.9540 2.263

Number of data N = 6, basic calculations on X and Y yield:

X  1.051/ 6  0.175 , S X  0.263 , Y  7.782 / 6  1.297 , SY  0.574 ,  X *Y  2.263
r
 XY  NXY 
2.263  6 * 0.175*1.297
 0.995
NS X SY 6 * 0.263* 0.574
SY 0.574
br  0.995  2.171
SX 0.263
A  log a  Y  bX  1.297  2.171* 0.175  0.9173  a  10 A  100.9173  8.265
Regression equation: y  axb  8.265 * x 2.171
Conclusion: Since the r value is greater, hyperbolic function is more reliable.
 79

Example 7.10: The relation between discharge (Q, m3/s) and water depth (h, m) (rating curve) in a
stream is Q  ahb . Obtain the regression equation considering the following data. Calculate Q value for h =
5 m and h value for Q = 100 m3/s.
Solution: The type of the regression equation Q  ahb is hyperbolic function:  y  ax b , so the equation
is made linear by using the following transformation: log Q  log a  b log h
Let Y  log Q , A  log a and X  log h , the equation is transformed to linear: Y  A  bX


GIVEN h (m) 1.25 1.50 2.00 2.50 3.00 3.50 -
DATA Q (m3/s) 4 7 13 25 44 70 -
SOLU-
TION: