Module 4

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Chapter 4

STATISTICAL TOOLS
The credibility of the conclusions of a research undertaking greatly depends on
the statistical tools employed in the analysis of research data. There are a lot of
statistical tools developed, but it is the “nature of the data” that governs the method that
is appropriate to interpret the data and the statistical tool that is required to process it.

In agricultural engineering, and land and water resources research, which usually
deal with numerical data, statistics can assist the researcher by indicating the central
point or the spread of the data and by showing the relationship between the
independent and dependent variables.

DESCRIPTIVE STATISTICS

These statistics describe the general characteristics of the variables, objects,


people or events under investigation. The most common descriptive statistical tools are:

1. Frequency Counts (f). This is the number of counts of a data of a given magnitude
or range in magnitude in a data set.

2. Percentage (%). This is the ratio of the frequency count of a data of a given
magnitude or range in magnitude to the total number of data in the data set
multiplied by 100, or in equation form:

Percentage = (f/N) 100

where “f” is the frequency count of a certain data of given magnitude or range in
magnitude and “N” is the total number of data in the data set.

3. Ratio. This is the quotient of dividing two quantities (a  b) or a fraction (a/b).

4. Arithmetic Mean (Xm). This is the most commonly used measure of central
tendency, which is computed by the formula:

5. Standard Deviation (s). This is the most commonly used measure of dispersion.
This is computed by the formulas:

INFERENTIAL STATISTICS

These statistics are used to test relationships, differences and interactions for
their significance. The test procedures employed on this kind of statistics are
categorized into parametric tests and nonparametric tests.

Parametric Tests

These tests are generally used when the variables to be tested are assumed to
be normally distributed and when the data from an engineering experiment are

26
measurement data (interval or ratio type). Some examples of measurement data in
engineering experiments are machine capacities, crop yields, soil and moisture losses,
flexural and compressive strength of engineering materials, hydrometeorological data,
electrical resistance, voltage, power consumption, and the like.

The most commonly used parametric tests are the following:

1. Student’s “t” Test. A t-Test is any statistical hypothesis test in which the test
statistic has a Student's t distribution if the null hypothesis is true. It is applied when
sample sizes are small enough that using an assumption of normality and the
associated z-test leads to incorrect inference. This is used for testing the difference
of two groups/samples when N is less than 30.

The t statistic was introduced by William Sealy Gosset for cheaply monitoring
the quality of beer brews ("Student" was his pen name). Gosset was a statistician
for the Guinness brewery in Dublin, Ireland, and was hired due to Claude Guinness's
innovative policy of recruiting the best graduates from Oxford and Cambridge to
apply biochemistry and statistics to Guinness' industrial processes. Gosset
published the t test in Biometrika in 1908, but was forced to use a pen name by his
employer who regarded the fact that they were using statistics as a trade secret. In
fact, Gosset's identity was known not only to fellow statisticians but to his employer
— the company insisted on the pseudonym so that it could turn a blind eye to the
breach of its rules.

In the use of the t-Test, the following assumptions are made:

a. Normal distribution of data


b. Equality of variances
c. Samples may be independent or dependent, depending on the hypothesis
and the type of samples. Independent samples are usually two randomly
selected groups whereas dependent samples are either two groups
matched on some variable (for example, age) or are the same people
being tested twice (called repeated measures)

The most frequently used t Tests are:

a. A test of the null hypothesis that the means of two normally distributed
populations are equal. Given two data sets, each characterized by its
mean, standard deviation and number of data points, a “t test” can be used
to determine whether the means are distinct, provided that the underlying
distributions can be assumed to be normal. All such tests are usually
called Student's t tests, though strictly speaking that name should only be
used if the variances of the two populations are also assumed to be equal.
There are different versions of the t test depending on whether the two
samples are:

i. independent of each other (e.g., individuals randomly assigned into two


groups), or

ii. paired, so that each member of one sample has a unique relationship
with a particular member of the other sample (e.g., the same people

27
measured before and after an intervention, or IQ test scores of a
husband and wife).

b. A test of whether the mean of a normally distributed population has a


value specified in a null hypothesis.

c. A test of whether the slope of a regression line differs significantly from


zero (0).

2. Z Test. This statistical procedure uses the normal curve to test the difference
between the sample mean and population mean when the population standard
deviation is known or to test the difference between means of two samples when the
sample standard deviations are known. A basic requirement for this test is that the
sample size (n) is 30 or more.

In testing the difference between the sample mean (xm) and the population
mean (Xm), the normal deviate (xd) is computed by the equation:
xd = Xm + Z S

where S is the population standard deviation. At 1% level of significance, Z = 2.58.


When xd > xm, the difference is insignificant; otherwise, the difference becomes
significant.

In testing the difference between two sample means (xm1, xm2) when the
samples standard deviations (s1, s2) are given, the normal deviate (xd) is computed
by the equation:

xd = Z (s12 + s22)1/2

At 1% level of significance, Z = 2.58. When xd > (xm1 – xm2), the difference is


insignificant; otherwise, the difference becomes significant.

3. F Test (Analysis of Variance or ANOVA). In statistics, the analysis of variance


(ANOVA) is a collection of statistical models, and their associated procedures, in
which the observed variance is partitioned into components due to different
explanatory variables. The initial techniques of the analysis of variance were
developed by the statistician and geneticist, R. A. Fisher, in the 1920s and 1930s,
and is sometimes known as Fisher's ANOVA or Fisher's analysis of variance, due
to the use of Fisher's F-distribution as part of the test of statistical significance. This
is used for testing the differences of the means of three or more groups.

The ANOVA is based on the following assumptions:

a. Independence of cases - this is a requirement of the design.


b. Normality - the distributions in each of the groups are normal
c. Homogeneity of variances - the variance of data in groups should be the
same

In practice, there are several types of ANOVA depending on the number of


treatments and the way they are applied to the subjects in the experiment:

28
a. One-way ANOVA is used to test for differences among two or more
independent groups. Typically, however, the One-way ANOVA is used to
test for differences among three or more groups, with the two-group case
relegated to the T-test (Gossett, 1908), which is a special case of the
ANOVA.

b. One-way ANOVA for repeated measures is used when the subjects are
subjected to repeated measures; this means that the same subjects are
used for each treatment. Note that this method can be subject to carry-
over effects.

c. Factorial ANOVA is used when the experimenter wants to study the


effects of two or more treatment variables. The most commonly used type
of factorial ANOVA is the 2×2 (read as “two by two”) design, where there
are two independent variables and each variable has two levels or distinct
values. Factorial ANOVA can also be multi-level such as 3×3, etc. or
higher order such as 2×2×2, etc. but analyses with higher numbers of
factors are rarely done because the calculations are lengthy and the
results are hard to interpret.

d. When one wishes to test two or more independent groups subjecting the
subjects to repeated measures, one may perform a factorial mixed-
design ANOVA, in which one factor is independent and the other is
repeated measures.

e. Multivariate analysis of variance (MANOVA) is used when there is more


than one dependent variable.

4. Regression Analysis. This is a technique used for the modeling and analysis of
numerical data consisting of values of a dependent variable (response variable) and
of one or more independent variables (explanatory variables). The dependent
variable in the regression equation is modeled as a function of the independent
variables, corresponding parameters ("constants"), and an error term. The error
term represents unexplained variation in the dependent variable. The parameters
are estimated so as to give a "best fit" of the data. Most commonly, the best fit is
evaluated by using the least squares method, but other criteria have also been used.

Regression can be used for prediction (including forecasting of time-series


data), inference, hypothesis testing, and modeling of causal relationships.
Regression models predict a value of the y variable using the given known values of
the x variables. If the prediction is to be done within the range of values of the x
variables used to construct the model this is known as interpolation. Prediction
outside the range of the data used to construct the model is known as extrapolation
and it is more risky.

The uses of regression rely heavily on the underlying assumptions being


satisfied, as follows:

a. The sample must be representative of the population for the inference


prediction.

29
b. The dependent variable is subject to error. This error is assumed to be a
random variable, with a mean of zero. On the other hand, the independent
variable is error-free.
c. The predictors must be linearly independent, i.e. it must not be possible to
express any predictor as a linear combination of the others.
d. The errors are uncorrelated and they follow a normal distribution. If non-
normal, the generalized linear model should be used. Furthermore, the
variance of the error is constant.

Regression analyses may be done in two ways:

a. Linear Regression Analysis. In linear regression analysis, a linear


relationship is said to exist between the dependent and independent
variable. The relationship between any two variables is linear if the
change is constant throughout the whole range of values under
consideration.

i. Simple Linear Regression Analysis – a linear regression analysis that


relates a dependent variable (y) to the value of one independent
variable (x) only by the straight line:

y= +βx

wherein  is the intercept of the line on the y-axis and β, the linear
regression coefficient, is the slope of the line or the amount of change
in “y” for each unit of change in “x”.

With the two parameters ( and ) of the linear relationship


specified, the value of the dependent variable “y”, corresponding to a
given value of the independent variable “x” within the range of “x”
values considered, can be immediately determined by replacing “x” in
the equation with the desired value and computing for “y”.

ii. Multiple Linear Regression Analysis – a linear regression analysis


exists when all independent variables are assumed to affect the
dependent variable in a linear fashion and independently of one
another. A multiple linear regression is said to be operating if the
relationship of

● a dependent variable (y) to the values of two or more independent


variables (x1, x2, …, xn) can be expressed:

y =  + β1 x1 + 2 x2 + … +  n xn

wherein  is the intercept (i.e., the value of “y” when all the “x’s” are
zeroes) and β1, 2, …,  n, the partial regression coefficient
associated with the independent variables “x1, x2, …, xn”, represent
the amounts of change in “y” for each unit change in “x1, x2, …, xn”,
respectively.

30
● a dependent variable (y) to the values of two or more functions of
the independent variable (x) by the straight line:

y =  + β1 f[x]1 + 2 f[x]2 + … +  n f[x]n


wherein  is the intercept (i.e., the value of “y” when all the “f[x]’s”
are zeroes) and β1, 2, …, n, the partial regression coefficient
associated with the independent variables “f[x]1, f[x]2, …, f[x]n”,
represent the amounts of change in “y” for each unit change in
“f[x]1, f[x]2, …, f[x]n”, respectively.

b. Nonlinear Regression Analysis. In this type of regression analysis, a


nonlinear relationship is said to exist between the dependent and
independent variables. The functional relationship between variables is
nonlinear if the rate of change in the dependent variable associated with a
unit change in independent variable/s is not constant over a specified
range of the values of the independent variable/s.

i. Simple Nonlinear Regression Analysis – In this analysis, a nonlinear


relationship exists between a dependent variable and an independent
variable. However, a simple nonlinear relationship (usually between
two variables) can be linearized by the following ways:

● Transformation of Variables – this is done by transforming one or


both of the variables through inverse, logarithmic and other
mathematical methods. Some common forms of nonlinear
relationships and their corresponding linearized form are given in
the table below:

LINEARIZED FORM THROUGH


NONLINEAR FORM
VARIABLE TRANSFORMATION
y = a ebx y' =  +  x
where: y’ = ln y,  = ln a,  = b
y = a bx y' =  +  x
where: y’ = ln y,  = ln a,  = ln b
1/y = a + b x y' =  +  x
where: y’ = 1/y,  = a,  = b
y = a + b/x y =  +  x’
where:  = a,  = b, x’ = 1/x
y = (a + b/x)-1 y' =  +  x’
where: y’ = 1/y,  = a,  = b, x’ = 1/x

After the linearization through variable transformation, the simple


linear regression procedure can be directly applied.

● Creation of New Variable – this is done by the creation of one or


more variables which can account for the nonlinear component of
the original equation. In agricultural engineering research, this
technique is most commonly applied to the nth degree polynomial:

y =  +  1 x +  2 x2 + … + n xn

31
which can be linearized by creating “n” new variables: z1, z2, …, zn,
to form a multiple linear equation of the form:

y =  +  1 z1 + 2 z2 + … +  n zn
where z1 = x, z2 = x2, …, and zn = xn. With this linearized form,
the procedure for multiple linear regression can be directly applied.

ii. Multiple Nonlinear Regression Analysis. When the relationship


between the dependent variable “y” and the “n” independent variables
x1, x2, …, xn, where n >1, does not follow the multiple linear
relationship, a multiple nonlinear relationship exists. However,
agricultural engineering research in the undergraduate level is unlikely
to go into this kind of analysis; hence, this is not given emphasis in this
manual. Rather, anyone interested in this type of analysis is being
referred to books in statistics.

5. Correlation Coefficient. This statistical parameter indicates the strength and


direction of a linear relationship between two random variables. In general statistical
usage, correlation or co-relation refers to the departure of two variables from
independence. In this broad sense there are several coefficients, measuring the
degree of correlation, adapted to the nature of data.

A number of different coefficients are used for different situations. The best
known is the Pearson product-moment correlation coefficient (PMCC), which is
obtained by dividing the covariance of the two variables by the product of their
standard deviations. Despite its name, it was first introduced by Francis Galton. The
Pearson PMCC is also known as the "sample correlation coefficient".

When measured in a population the Pearson PMCC is designated by the


Greek letter “rho (ρ)”. When computed in a sample, it is designated by the letter “r”
and is sometimes called "Pearson's r." Pearson's correlation reflects the degree of
linear relationship between two variables. It ranges from +1 to -1. A correlation of
+1 means that there is a perfect positive linear relationship between variables. A
correlation of -1 means that there is a perfect negative linear relationship between
variables. A correlation of 0 means there is no linear relationship between the two
variables. Correlations are rarely if ever 0, 1, or -1. A certain outcome could indicate
whether correlations are negative or positive.

The linear equation that best describes the relationship between “x” and “y”
can be found by linear regression. This equation can be used to "predict" the
value of one measurement from knowledge of the other. That is, for each value
of “x” the equation calculates a value which is the best estimate of the values of
“y” corresponding the specific value.

6. Least Significant Difference Test (LSD Test). This is the simplest and the most
commonly used procedure for making pair comparisons. This procedure provides
for a single LSD value, at a prescribed level of significance, which serves as the
boundary between significant and non-significant differences between any pair of
treatment means. That is, two treatments are declared significantly different at a

32
prescribed level of significance if their difference exceeds the computed LSD value;
otherwise, they are not significantly different.

The LSD test is most appropriate for making planned pair comparisons but,
strictly speaking, is not valid for comparing all possible pairs of means, especially
when the number of treatments is large. Thus, avoid the use of the LSD test for
comparison of all possible pairs of means. However, if this must be used, apply it
only when the F test for treatment effect is significant and the number of treatments
is not too large – less than six.

The procedure for the LSD test varies from one experimental design to
another. Please refer to the book entitled “Statistical Procedures for Agricultural
Research (Second Edition) by Gomez and Gomez” for the detailed procedure as this
is not presented in this manual since this is quite lengthy.

7. Duncan’s Multiple Range Test (DMRT). This the procedure most applicable in
making comparison between all possible pairs of treatment means, especially when
the total number of treatments is large.

The procedure for applying this test is similar to that of the LSD test. This
involves the computations of numerical boundaries that allow for the classification of
the difference between any two treatment means as significant or non-significant.
However, unlike the LSD test in which only a single value is required for any pair
comparison at a prescribed level of significance, the DMT requires computation of a
series of values, each corresponding to a specific set of pair comparisons.

Just like the LSD test, the procedure for the DMRT varies from one
experimental design to another. Please refer to the book entitled “Statistical
Procedures for Agricultural Research (Second Edition) by Gomez and Gomez” for
the detailed procedure as this is not presented in this manual since this is quite
lengthy.

Nonparametric Tests

These are tests generally used when the distribution of the variables to be tested
is assumed to be non-normal and when the data from an agricultural engineering
research study are attribute data (nominal or ordinal type). The most common types of
attribute data are those having two classes, which consist of the presence or absence of
an attribute, such as success or failure, effective or ineffective, mechanized or manual,
dry or wet, high or low, and the like. Examples of attribute data with more than two
classes are mechanization status (full, partial, none), moisture status (flooded,
saturated, moist), grain quality (whole, cracked, broken), size of gravel particles (large,
medium, or small), level of resistance (high, medium, or low), and the like. Furthermore,
the effectivity of a soil erosion control measure may be scored on a scale of 0 to 5
instead of measuring the actual amount of soil loss from the erosion plots.

The advantages of nonparametric over the parametric tests are:

a. Nonparametric tests can be correctly applied to a much larger class of


population than the parametric tests because they are based on fewer and
less number of assumptions.

33
b. Most non-parametric tests are generally non-mathematical and are faster to
do than the parametric tests since the former are based on ranking, counting,
addition and subtraction.
c. When the sample sizes are small or moderate, non-parametric tests are faster
than the parametric tests.
d. Non-parametric tests usually require data, at least on the ordinal scale,
though sometimes nominal data can also be used. Parametric tests, on the
other hand, generally require measurements on the interval or ratio scale.
e. As assumptions are fewer with non-parametric tests, they are less susceptible
to violations.
f. When sample sizes are small, say 10 or so, the distribution-free (parametric)
statistics are easier, quicker and only slightly less efficient as compared to
their classical (parametric) equivalents.

On the other hand, the disadvantages of non-parametric tests are:

a. When dealing with large sample size, non-parametric tests are more time-
consuming and tedious, and, hence, become much less efficient compared to
the parametric tests.
b. Non-parametric tests waste information.
c. They tend to lead to the acceptance of the null hypothesis more often than
they should.
d. They are less efficient than the parametric tests.

The most common nonparametric tests employed in agricultural engineering


research and development are the following:

1. Chi-Square Test. A chi-square test (also chi-squared or χ2 test) is any statistical


hypothesis test in which the test statistic has a chi-square distribution when the null
hypothesis is true, or any in which the probability distribution of the test statistic
(assuming the null hypothesis is true) can be made to approximate a chi-square
distribution as closely as desired by making the sample size large enough. This is
considered a versatile test since it can be used to determine significant association
or dependence as well as to test for significant differences.

The chi-square test is a powerful statistical tool widely applicable for the
analyses of geographical data which are mostly nominal and are based on some sort
of frequency groupings. Introduced in 1900 by Karl Person, this test is the most
important of all distribution-free (non-parametric) tests.

This test is made possible by calculating the value of a quantity called chi-
square (2) which is dependent on the observations between observed and expected
class totals. Thus, the test is a simple technique which works by comparing the
actual frequencies observed with a regular frequency expected.

For the application of the chi-square test, the following requirements should
be satisfied in the data:

a. The data must be in the form of ordinal data and it should be based on the
frequencies or counts obtained in each number of categories. In the data,
percentages or rates per thousand cannot be used.

34
b. There must be at least two mutually exclusive categories into which the
observations are placed and they must be independent and random.
c. The total numbers observed must exceed 20 and preferably greater than
40.
d. The expected frequency in any one category for a 2 x 2 problem (2 rows
by 2 columns) must normally not be less than five. However, if there are
five or more categories, then not more than 20% of the expected
frequencies may be less than five and there should be no category with an
expected frequency less than one. It should be noted that this rule is only
for expected frequencies and not for the frequencies actually obtained in
the data which can be of any size.

The chi-square test has the following advantages:

a. This test enables one to compare the frequency distributions rather than
mean values which most parametric tests can do.
b. As compared to the “t Test”, the chi-square test considers the whole range
of data and not merely the mean and standard deviation.

Some examples of chi-square tests applied in agricultural engineering


research include:

a. Test for goodness-of-fit – determines whether a set of observed data


conforms to a specified probability distribution (for example: the number of
agricultural machineries owned by farmers may be suspected to follow a
normal distribution)

b. Test for independence – determines whether two sets of observed data


are independent from each other (for example: ownership of agricultural
machineries and geographical area).

2. Mann-Whitney U Test. The Mann-Whitney U test (also called the Mann-Whitney-


Wilcoxon [MWW], Wilcoxon rank-sum test, or Wilcoxon-Mann-Whitney test) is a
non-parametric test for assessing whether two samples of observations come from
the same distribution or to test the difference of the means of two groups. The null
hypothesis is that the two samples are drawn from a single population, and therefore
that their probability distributions are equal. It requires the two samples to be
independent, and the observations to be ordinal or continuous measurements, i.e.
one can at least say, of any two observations, which is the greater. In a less general
formulation, the Wilcoxon-Mann-Whitney two-sample test may be thought of as
testing the null hypothesis that the probability of an observation from one population
exceeding an observation from the second population is 0.5.

This test is one of the best-known non-parametric significance tests. It was


proposed initially by Wilcoxon (1945), for equal sample sizes, and extended to
arbitrary sample sizes and in other ways by Mann and Whitney (1947). MWW is
virtually identical to performing an ordinary parametric two-sample “t test” on the data
after ranking over the combined samples.

35
The U test is often recommended for situations where the distributions of the
two samples are very different.

3. Kruskal-Wallis H Test. This is the direct generalization of the two-sample


Wilcoxon-ranked pair test. This is an appropriate test whenever there are more than
two independent samples and an ordinal scale level of measurement, and there is a
need to assess the contribution made by each of them in the total variability of data.
This is used to test whether or not a group of independent samples is from the same
or different populations, or for testing the difference of the means of three or more
groups. The Kruskal-Wallis test is a nonparametric alternative which does not rely
on an assumption of normality.

This test requires that all the observations under the samples are ranked
together from the smallest to the largest. The sums of the ranks are then computed
for each of the nominal scale categories of the variables and they are designated as
R1, R2, …, Rn. Then, a test statistic H is computed in order to measure the degree to
which the various sum of ranks differ from what could be expected under the null
hypothesis.

However, the limitation of this Kruskal-Wallis Test is that it does not isolate the
pairs that have different means nor, by accepting the null hypothesis, it does not
exclude the possibility that for one pair of samples the null hypothesis could be
rejected by the Wilcoxon test at the same significance level.

36

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy