RESEARCH METHODOLOGY

Yoga Research:

1. Revalidate empirically
2. Ancient literature
3. Subjective and objective
4. Multi disciplinary

Statistics is a tool when we have different effects it is helpful to express.

Statistics and RM concepts are two wings to fly high.

Qualities of a Yoga Researcher:

1. One must be a practitioner of Yoga.

Qualities of a Good Researcher:

1. Good observer
2. Unbiased
3. Curiosity
4. Perseverance
5. Good Communication
6. Scientific Temper
7. Knowledge of the domain
8. Ethics
9. Creative
10.Vigilance / Alertness
11.Planning ability
12.Discipline

HW. Write the above 12 points and what you are in these respects. How can you
improve yourself?

Research process: ( Process involves series of points and the way of doing)

1. Research Question
2. Literature survey
3. Planning ( 1,2 and 3 together is Phase 1)
4. Data ( Phase 2 )
5. Data Analysis and interpretation
6. Conclusion
7. Communication ( 5, 6 and 7 together is Part 3)

1. Knowledge of RM – Phase 1
2. Field work – Phase 2 ( There is no course. Go and help seniors. Learn by
observation. Involve in field )
3. Knowledge of Statistics – Phase 3.

Phase 1 :
Coming out with foolproof methods of doing your research.
You need to think in diverse way. If you know all the ways your project can go
down.

Process of research Question:

Research Question:
What exactly you want to know- You should know
Question that tells about research
A precise statement
A statement need to be constructed

Qualities of a Good Research Question:

1. It should contain the problem.
2. It should be precise
3. Method of intervention should be clear
4. Population should be defined
5. Tools should be known
6. Should be short and precise
7. Variable
8. Expected outcome
9. It should be applicable to larger population
10.It should be socially relevant.
11.We should be able to adhere to the time line
 12.It should be realistic
13.It should be doable
14.It should add to knowledge
( It should advance existing knowledge.)

 Literature Review:
 Google Scholar
 Pubmed
 Science direct
 Sci_Hub (To open the full paper)
 Library genesis
 Li_Gen (To open the books)

21/8/17
Research Question:

1. It should be novel advancing the current knowledge

2. Socially relevant
3. Time bound
4. Precise

HW Take three statements of research questions and evaluate each one of them.

How do we ensure a research question is novel?

Look for previous research. Do Literature review.

You need to search whole track.

Focus on latest research

 Past 3 yrs
 Past 5yrs
 Past 15 yrs

Reproducible research – Transparency

Open science frame work

OSF

Helps to bring out reproducible research

Method of going to previous Research is called Literature review.

22/8/17

Literature Review:

To ascertain that the work we are going to do is Novel.

Why? ->

1. To know what is done?

2. To know what is needed to be done?
3. New ideas of Research Question
4. Give insight into methodological details.
5. Gives strengths and weaknesses of the earlier studies.
This helps to strengthen the possibly weak areas in that field.
6. Saves time and resources ( by giving an idea of what problems earlier
researchers faced.)
7. To take precautions.
8. Connect to people who are working in that regions.
( To know the researchers in your field.)

How and when Literature review has to be done?

Phase 1 – 80%
Phase 2: 5%
Phase 3: 15%
It is a continuous process.

24/8/17

How to do Literature Review?

1. Precisely write down what you want to review about?
2. Eg: Pranayama – Memory
3. Yoga – Intelligence
Scientific Journal is a collection of Scientific Articles.

It is a collection of recent research being done in a particular field.

1. Define topic of Literature Review.

2. Sources of information.
Library – scientific journal.
3. Select the key words
4. Organize information ( literature review that is collected.
5. Synthesize

Thesis is advanced

PhD Thesis

MSc Dissertations

Thesis

News Paper – Current emerging trends will be known to decide topic.

TV

Journals or dissertation is the primary source of information. Anything which is pear

reviewed and validated.

Secondary source of information is books, magazines, TV and news paper.

UGC has a central repository of all dissertations collected as Shodganga.

Online – Digital

Off line - Printed -Library

Online Scientific data bases:

Pubmed
Science direct – Science Engg

Psychinfo ( psychology Education)

ERIC

Shodhganga

26/8/2017

Keyword – Yoga, Memory,

Pranayama, Memory

Right nostril Breathing. Memory

Read one article per wk

After 4 wks

Read 5 articles per wk

Suppose you downloaded 500 articles

Which one to read 1st?

Start from latest

 1yr
 3yr
 5yr
 10yr
 15yr

Classic research – 1st time done

Journal – Original articles where you present your experimental results

- Review articles ( synthesis of already existing research )

- Single case studies ( Medically unusual cases )
 - Correspondence

Impact Factor of the Journal

Hindex – Individual research Articles are cited how many times?

Social relevance: Tutor – Science Research to common man

Review Articles can be of three types. They are:

1. Narrative review: An experienced author gives it gives it depending on his life

time experience in that field of study.
2. Systematic Review: You put certain criteria and present that which satisfy that
criteria.
3. Meta Analysis: Effect sizes of various articles are collated and a common Effect
Size is found.
Meta analysis has all qualities of systematic analysis and also collates all
statistics of all papers.
Strength of Evidence is most in Meta analysis and then in systematic review and
then narrative.

Mendeley and Decear help in Arranging or organizing Literature.

A summary table of the reviewed literature need to be prepared as per the
interest of the person. For example summary table may have the following
columns.
Author, Year, Etc.

Edmodo for students

Dzurgr
Pubmed.com
Sciencedirect.com
Free PMC article – Full text article can be got without paying
Library genesis/Libgen
Research methodology
Statistical Analysis
Name of book
On line courses
Courser – Free

31/8/17
What is the plan/ Scheme/ Logic that we are going to check need to be
decided.
More time need to be given for planning

NULL HYPOTHESIS SIGNIFICANCE TESTING (NHST):

Hypothesis is an anticipated statement about possible outcome of an

experiment.

Practice of Pranayama changes in memory

Practice of Pranayama indicates no change in memory

Inductive logic Deductive logic

If you do not know anything, check whether it is reoccurring then make

hypothesis. This is called exploratory research.

99% of MSc research is experimental research

Null hypothesis is a statement which negates statement of Hypothesis.

Null Hypothesis is most important in whole process of research.

According to Western Law in order to establish a truth it is not sufficient

to say:

Something exists.

It is also required that

Non-existence of “something” does not exists also to make it stronger.

 Eg: I see smoke.

I infer fire is there.

It is responsibility of researcher to prove to 90% possibility at least about

his inference about existence of fire.

Eg: I see sky in day time,

I conclude saying there are no stars in sky.

You can have of number of hypothesis for a Hypothesis. You can keep
negating them one by one.

1/9/2017

When we try to see hypothesis, then only we may see only one aspect of
truth and may lose to see all dimensions. All Null Hypothesis have to be
negated to say that Hypothesis is correct.

Most of the researches are of inferential nature.

By negating Null Hypothesis we get support to Hypothesis.

Eg: Out of 10 Null Hypothesis, we get 7 Null Hypothesis is negated. I.e.

we get 70% support to Hypothesis.

Hypothesis: Practice of pranayama increases memory.

NULL HYPOTHESIS:

1. Practice of pranayama does not increase verbal memory.

2. Practice of pranayama does not increase spatial memory.
(While writing Null Hypothesis invariably you need to attach the
variable explicitly explained.)
3. Practice of pranayama does not increase long term memory.
If 3 is not negated,
Under the given experimental conditions long term memory is not
increased.
The more the number of null Hypothesis you reject, stronger is the
evidence.
 As a researcher, you should not feel bad by seeking negative results.

Research Hypothesis / Alternative Hypothesis/ Ha

When researcher does not speak about the direction of change then
it is called non-directional hypothesis.

Steps in NHST:
1. State Alternative Hypothesis
2. State Null Hypothesis
3. Fix Alpha
4. Fix Power ( 1-Beta )
5. Estimate Effect Size.
6. Calculate required sample size
7. Collect data
8. Analyze and conclude ( If p < alpha, reject NH; if p > alpha Fail to
reject NH )

Directional Hypothesis is called one tailed Hypothesis.

Non-directional Hypothesis is called two tailed hypothesis.

When researcher speaks about the direction of change then it is called directional
Hypothesis.

7/9/17

1. Ha
2. Ho
3. Fix α
4. Fix power (1-β)
5. Estimate Effect size
6. Calculate Sample Size
7. Collect data
8. Analyze and conclude - If p<0.05 => Ho rejected
- If p>= 0.05 => Fail to reject Ho
Understanding α,β, power, effect size, sample size

Whenever we make an inference, there is always a possibility of error ( or chance of

occurrence of error)

Two types of error –

1. α error (Type 1)
2. β error (type 2)

eg. Ha : Bhramari Pranayama (Bh. Pr.) improves memory

H0 : Bh. Pr. Does not improve short-term memory

Conditional probability

Sum of the probabilities is always 1.

Decision:

H0 True H0 False
Reject H0 × 
Type 1 / α 1-β

Fail to reject  ×
H0
1-α Type 2 / β

1 – β is known as POWER.

Suppose H0 is true

i.e. Bh. Pr. Does not improve short-term memory

Suppose H0 is false

i.e. Bh. Pr. Improves short-term memory

Reject H0 :
Researcher claims that Bh. Pr. Improves short term memory

Fail to reject H0 :

Researcher claims that Bh. Pr. Does not improve short-term memory.

α is traditionally fixed as

- 0.05 - *
- 0.01 - **
- 0.001 -***

I can tolerate the error maximum up to 5% only then we fix α = 0.05.

Fixing power is the level at which you claim it is correct when it is correct.

Power traditionally is fixed as

- 0.80
- 0.95
- 0.99

8th September 2017

Since α error appears to be more serious, we need to fix α first.

We control β error by fixing power.

Any overestimation of experimental result leads to α error.

Any underestimation of experimental result leads to β error.

Eg. An average class student ability is 75%

- Paper is very difficult so students get 30% and seeing that, I assume
students are weak. This is underestimation that leads to β error.
- Paper is very easy so students get 90% and seeing that I assume students are
very smart. This is overestimation that leads to α error.

Effect size, sample size :

Before we start our research, we need to have required sample size known.

Size of the effect is called effect size.

Magnitude of intervention is called effect size.

1 day Yoga – Effect size

10 day Yoga – Effect size

100 day Yoga – Effect size

Before experiment starts, how can we estimate effect size?

1. By doing Literature Review

- Refined definition of effect size is the magnitude of change due to the
intervention as measured by your measuring variable.
- I should choose a research with similar intervention, similar measuring
variable for using literature review for effect size.
2. By doing Pilot study
- Pilot study with a small scale study with similar intervention similar
measuring variable when literature is not available.
3. Standards

Effect size is defined as – High, Normal, Low

Cohen – D

Compare mean Correlational

High 0.80 0.5
Normal 0.50 0.3
Low 0.30 0.1
21/9/17

Assignments given on Edmodo, deadline given

Test1 announced

Sensitivity of the measuring tool.

Fix α, Fix 1-β, Estimate ES are done to calculate the required Sample Size.

If you don’t have enough SS, sometimes you may end up with inconclusive results.

Apriori power analysis – Before data collection

Post hoc power analysis - After data collection

If power < 80% and p>0.05 then we can surely say intervention is not useful.

Apriori Power analysis:

Before data collection

To calculate the required Sample Size.

Posthoc power analysis:

After Data collection

To calculate the achieved power

Apriori:

SS ->α

 1-β
 ES
 Tail

Post hoc:

1-β -> α

 SS
 ES
 Tail

22/9/17

Longer the intervention – Chance of drop out is more.

Literature usually gives an idea about dropout %

After study how much power you could achieve? Is it 80% or more or not.

P-Value(probability)

The value of probability ranges between 0 – 1

0 – Never

1 – always

Above 0 and Less than 1 is the range.

Stringent the constraint, SS will be higher.

Lower α more SS

More power ( 1-β ) more SS

Small ES more SS

Rejecting a 2 tail hypothesis – More SS

P(D/Ho)

The probability of change happening by random chance is P value.

When p< 0.05 always reject Ho

p>= 0.05 Fail to reject Ho

p value tells the experimental change assuming the Null Hypothesis to be True.

23/9/17
P value is the probability of finding the experimental change assuming the Null
Hypothesis to be True.

*Conditional probability is the probability of finding the experimental change by

random chance assuming the Null Hypothesis to be True.

5/10/17

Test papers giving:

P = 0.082

1-β= 0.75

The study is inconclusive

Post hoc power = 0.65

He decides to conclude that intervention does not work.

29/9/17

STUDY DESIGN:

Planning is all about designing and experiments

Philosophy of design.

The happening of an event

Signal – Information which we actually desiring to capture

Noise – Unwanted ( which we don’t want ) we always try to enhance the signal and
reduce the noise.

Signal/Noise

If we are not able to enhance signal and try that the ratio do not increase
If we want to increase the signal, directly work on intervention.

Eg: 1 WK to 4WK

Confounding factors are disturbing factors.

Researcher must try to minimize confounding factors.

Secure the signal – Minimise the Noise

If we minimize the noise/ disturbance then our signal can be more.

Eg: Singer in Jigini Bus Station -> Noise

Singer in SVYASA Music class -> Signal

Noise enhances our errors -> α and β errors.

Variables : confounding factors

Variables are not constant. It changes from person to person. Variables are not
constant. It changes from person to person.

Variable is something which varies which has scores.

Types of Variables:

1. Independent ( cause ) – Fixed

- Measured
2. Dependent ( effect )
3. Confounding

Cause(Yoga ) -> Effect( Memory)

Independent variable is related to cause. It is manipulated/ controlled by

researcher.

Eg: Duration of intervention, but I choose to give one wk intervention.

Interested in studying – Independent variable

Characteristics of IV:

Tell something about cause.

It is controlled by researcher.

Dependent variable: Depends on independent variables. It is related to effect.

Variables which tells about effects of intervention.

Characteristics of Dependent variables:

Talks about effects

It is dependent on researcher.

 Confounding variables: are influencing the dependent variables but the

researcher is not interested in studying that.

Characteristics of confounding variables:

 It is influencing the dependent variables.

 Not interested in studying confounding variables.

1. List out all the confounding factors

 In what all way
 By which
 My experiment may go wrong
 Diet
 Memory Training
 Age, weight

2. Give it a priority
 High
 Low
 Moderate
3. Develops strategies
 Experimental Method – Actually do to control it
  Statistical method – We must have measured it to do this method.
 Statistical method – directly connected with study design.

5/10/17

Population
Population
All people who are eligible to participate in your study.

Sample: is a subset of a population.

Sample
Sample is a small group of people taken from population.

The process of selecting sample from population is called as Sampling.

Sample that you choose should be a representation of the population.

There are two types of Sampling techniques.

1. Probability Sampling
2. Non-probability Sampling.

1. Probability Sampling: There are four types of probability Sampling.

a) Simple Random Sampling.
b) Systematic Random Sampling
c) Stratified Random Sampling
d) Cluster Random Sampling.

2. Non-probability Sampling: There are four types of probability Sampling.

a)Convenience Sampling
b)Convenience Sampling
c)Snow ball Sampling
d)Quota Sampling.

6/10/7

Simple Random Sample:

This is rarely done when population is big.

Prepare the list of all eligible participants.
Randomly select required numbers.
(randomisor.org will be used
Lottery
Random number table
Ask computer for random numbers.)

Systematic random sampling

We make list of all eligible participants .We choose (select) systematically.

Eg. 1, 7, 17, 21, …. Etc

Stratified Sampling: This is applied when you know the characteristics of the
population well in advance, you must see that the ratio is represented in the sample.

Eg: For the disease of study women are more prone than the men, then you need to
take a sample with 75% women and 25% men.

Cluster Random Sampling:

When we need to cover large Geographic area, this is done. We make clusters in
geographical region and randomly select clusters from them.

You may adopt multiple sampling in your research.

Non-Probability Sampling:
You need not have to ensure that sample is a true representative of population.

You can generalize the results

Convenience Sampling:

Whoever is ready to join your research, take them.

Snowball Sampling: Word of mouth.

Eg: Someone comes to Arogyadhama, gets well, he goes back and tells another –
another one will come etc

Socially sensitive issues will be tackled in this way.

Quota Sampling:

Prepare a quota, as that no. of requirement is finished, you stop taking them.
Eg: I need 20 persons with HT with retinopathy
30 persons with HT and Head ach.

Purposive Sampling:
To experiment on Monks

We go to them each one individually and do.

Multistage sampling is one where we apply multiple types of samplings.

Generalisation: the process after doing Randomized sample, can be applied for
whole population.

Methods of selecting a Sampling for a specific research study.

Research Designs:
1. Experimental Design: You try to manipulate Independent Variable and try to
see its exclusive effect on Dependent Variable. Randomization is there.
Random allocation of the participants in different experimental conditions.
Confounding variables are 100% controlled.
2. Non-Experimental Design: You are not interested in studying cause and
effect. Confounding Variables are not at all controlled.
3. Quasi Experimental Design: It is like an experimental design but differs in
some aspects. There is no randomization. Confounding Variables are partially
controlled.

Samplings select sample

Randomisation divides sample for (experimental and control groups) to different
experimental conditions.

Non-experimental : survey design , try to get various characteristics of sampling

Observational design: also called field observation. Go there, take a video, analyse, study.
Qualitative studies come under this. Sports Aggression: aggression involved while playing
game.

Cross-sectional study:

Eg. Study the difference in memory of Gurukula students& normal students. We try to
compare key characteristics of distinct populations.

Eg. Meditation, novices

7.10.2017

Quasi-Experimental designs

Simple pre-post test design

PRE POST

Control group:
The group which does not do intervention and continues their own trajectory of life is
called control group. Control group is supposed to control the confounding factors.
Control group controls confounding factors by equally distributing across both groups.

Types of control groups

1. Normal CG : does not do anything

2. Active CG : Which does not simply sit. Eg. Exercise group, actively involved in
something
3. Waitlist CG : ask them to wait till the end of the experiment then they will be
given intervention.

DESIGNS

NON- QUASI-
EXPERIMENTAL
EXPERIMENTAL EXPERIMENTAL

Y   Survey 
PRE POST PRE POST
C  Observational/
PRE 
POST PRE POST Y
 field
 Cross-sectional
PRE POST
C


ANALYSIS OF VARIANCE – ANOVA

Whenever there are more than 2 or more than 2 assessment points, that design is
called ANOVA.
e.g. Guna measured at every semester beginning. This is called repeated measure
ANOVA. It is a within group ANOVA because same subjects participate in differen
conditions.
Between Group ANOVA – different subjects – different conditions.

YOGA
AYURVEDA Between group ANOVA

CONTROL

MIXED DESIGN

PRE POST FOLLOWUP

Y Y Y

A A A

C C C

Two groups pre-post can be written or between groups.

When there is more than two groups or more number of assessments it is ANOVA

** Same subjects – Within

** Different subjects – Between

C C C

or

C
 C

or

C C C

C C C

It is ANOVA

Prospective ( For Future design )and Retrospective ( Look into records and carry
research ) Designs.

Cohart ( Means Group )

Like a crosssectional

Can be Quasi experimental

Longitudinal: Carried for long time ( 100 Yrs Etc)

12/10/17

Selection Bias: is more applicable to sampling. If you take a sample which is not
really representative.

Recall Bias:
Eg: Collecting your experiences of past one month and answer the questions.
Reporting may be influenced by our mood now.

We tend to distort the information while trying to recall information from the
memory.

Observer Bias:

In the process of observing, if we make any bias then it is called observer bias. It is
focused on the experimentor. Hence it is called experimentor bias.

Instrumental Bias:

Weighing machine not showing ‘o’ when nothing is there.

Instrument should be calibrated. Setting measuring instrument to zero in a non

measuring state.

Publication Bias:

Any bias introduced in the process of publication.

Fire drawer problem:

Publishing only positive results and not publishing negative results.

Problem with randomization is that we can not ensure all factors to be randomized.

Eg: When gender os important.

1st do matching then do randomization

Yoga Control

75M 25F 75M 25F

15M 15F 15M 15F

When you know confounding factors clearly, do randomization.

 CF

Known Not known

Experimental Statistical Experimental method Randomization

Inferior

ANCOVA ( Analysis of covariance)

Matching

Restriction

Blocking

Age Matching:

Yoga Control

20-30Yrs - 15 20-30yrs - 15

30-40yrs - 25 30-40Yrs - 25

40 – 50Yrs –10 40-50Yrs –10

Restrict: Certain group of people to participate in your study. This is done by exclusion
criteria.

Blocking: Allow certain people to restrict.

Eg: People with more age and are forgetful to be avoided.

Eg: Study is only for moderate level of memory.

Statistical method of controlling confounding variables is called ANCOVA. Initially data is taken
on all and extremes which are unwanted are controlled by statistics.

13/10/17

Order effect: Short time effect of Yoga practices.

Ie the order in which the practices are given effect the study.

If we want to see the effect of right nostril breathing and left nostril breathing possibly

We take 3 groups – 1. Control

2.Left nostril breathing

3. Right nostril breathing

This is between groups.

Whenever we study short time effect , the subject variability may lead to more
confounding.

Hence in such cases we need to avoid subject variability.

Withing group subject variability can be controlled effectively. If subject variability

is primary, choose within group design. Statistical significance purely goes with p
value ie p<0.05

Clinical significance looks at the change in the value. Clinical significance means
wheather the range changed.

BP

Low Normal High Very High

40 120 140 200

Jump in range

Pre______RN_______Post_____Washout period____Pre________Post

10Min Carry over effect 10Min

Washout period: The main period required to wash out the effect of earlier
intervention.

Wheather RN first and then LN next.

LN first and then RN next

The order in which you do is going to be confounding. It is called order effect.

Cross over design: Day1 Day2

RN 15

30

LN 15

Crossover design handles order effect.

Carrier effect handled by enough washout period.

Subjects don’t know wheather they are in control group or in experimental group. It is
called single blinding.

Placebo Effect: The psychological effect gets corrected. The participant takes pill
similar to the actual medicine and gets cured.

Drug trial group placebo group

Participant don’t know which group they are in.

Double Blinding: Experimentor also doesn’t know which group they belong to.

Tripple blinding: Neither participant, nor experimentor, nor statistician know which
is control, which is experimental.

Single case study: Try to study one subject in detail.

14/10/17

Measuring process:
Validity and Reliability:

Are the processes by which we can ensure quality of our research.

Ensuring the correctness of something is validity.

Being stable is reliability.

Validity:

1. Internal Validity
2. External validity
3. Construct validity
4. Statistical Validity
Are all quality measures.
1. Internal validity deals with the cause and effect relationship is appropriately
and correctly got.

Yoga ----- Memory ( Basic strength of the experiment )

Cause Effect

You can improve internal validity by Randomization.

External Validity: To expand the study results to various study results.

1 2

100 100

Bangalore Mumbai Generalisation

Random sampling/ Probability sampling improves external validity. Sample size must
be large.
Crossectional / Multicentric style:

Set up Set up Set up Set up

Mumbai Delhi Bangalore UP

A consept which is difficult to see. Intangible ( not seen directly ) but want to
measure.

Construct validity: ( Mental concept)

It is an abstract concept which is intangible but researcher still want to measure it.

Is the measuring tool calculations that which it is supposed to?

To ensure the construct which is measured by measuring tool…..

Eg: Depression Questionnie ,re really calculating the depression.

We should define the construct very well.

The process of define the construct is called operational definition. ( Op[erational

means dynamic. Definition means working definition).

Statistical Validity is when we choose appropriate test ( statistical ) test is statistical

validity. Statistical ( experimental design is fixed thus statistical tests).

Right test

Interpretation of number.

Analysis of data.
Reliability:
1. Test- Retest reliability
2. Split half Reliability
3. Internal Consistency
4. Inter Rater Reliability

Are the measuring procedures.

1. Test- Retest reliability: We measure group of people now and we retest and
see the correlation between the two.
The correlation between test scores and Retest scores.
Correlation when = Reliability

No intervention Invalid
50Kg 55Kg
The duration of Test Retest Reliability is chosen is based on the stability of construct.

Quality and Quality Tool: 1

1 2
2
2. Split half reliability tool: 3 3 4
Split the questionnier into two equal parts 4
5 5 6
Split half reliability = 1st part is not equal to other half
7 8
6
9 10
7

10

3.Internal Consistency:
 All the possible ways to devide into two groups. One ultimate way is called
(Chrombacks’ alpha ) α

1 2 3 4 5

Chronback ALPHA

Method of deriving reliability index by splitting the questionnier into all possible ways
and deriving a common Index.

4.Inter Rated Reliability:

High correlation between raters = Inter rated Reliability

Reliability between different Raters in giving openions

Population parameter
(The information which we get
from population)

Sample statistics
Generalisation (The information which we get from Sample)

 Statistics help in the process of generalization

 Statistics help us to derive information about sample
 Statistics help us to know us population parameters.

Statistics

Descriptive Inferential

Directly see and conclude Making some assumption after seeing the
situation
Example: How many samples are there in my research? Eg: Seeing the smoke we
assume that there
is fire

Descriptive statistics helps in describing the sample. Inferential statistics help in

describing the population.

Descriptive Sampling

Measure of central tendency Measures of Dispersion or

measures of spread

Tells about the central part Tells how much information is

scattered

How many signal is there How much noise is there

Measure of Central tendency:

1. Mean
2. Median one unbias number
3. Mode

Higher the frequency, there is more number of statistic data.

Helps in arriving one number which tells maximum story of the sample.

Example: CR will tell about student about an event

_
1. Mean ( X)(average) : Sum of all the numbers/ Number of Sample
Eg: 1, 2, 3, 4, 5 ( 1+2+3+4+5)/ 5=3 Mean
Mean is not always good to calculate.
Eg: 1,2,3,4,100 Extreme values / Outlayers

Mean should not be used when there are out layers in the data.
2. Median : Middle value after arranging the data in descending or ascending
order.
Eg: 1,2,3,25,5,7,75
 1,2,3,5,7,25,75
Median is 5
When we have outliers in the sample, Median gives best value for the result.
If we have two middle numbers, average of it has to be taken as median.
Eg: 1,2,3,100 ( 2+3)/2 =2.5 Median
3. Mode: Mode is the most frequently occurring number in sample.
Eg: 1,2,2,2,4,5,3,5,3

Unit no No of Times
1 1
2 3
3 2
4 1
5 2

More than two numbers can be there in sample then it is called Bimodes.

21/10/17
Measures of Dispersion:
1. Range
2. Standard Deviation
3. Coefficient of Variance
4. Variance
All these above four estimate the variability in the data.
1. RANGE: The difference between highest and the lowest score in the range.
One value
Eg: 1, 2, 3………75 75-1=74 range

X X bar (X-Xbar )2
1 -2 4
2 -1 1
3 0 0
4 1 1
5 2 4
Sigma 15 10 Sum of
squared deviations

Mean Xbar = 3
 (X-Xbar) = Deviance Gives us information of variance through
central tendency
Sigma (X-Xbar)2 /n = Variance 10/5 = 2 Variable

Sqrt(Sigma(X-Xbar)2 /(n -1 ) = Standard Deviation Sqrt(10/4) = Sqrt(2.5)

n-1 is the degrees of freedom.

Ie the number of scores in variables that can actually vary (n-1)

Use n-1 when we calculate sample descriptive statistics

N when we calculate population inferential statistics.

Coefficient of Varience:

With respect to mean what is the corresponding Standard deviation is called

coefficient of variance.

Coefficient of Variance(CV) = SD/Xbar

It is used to check reliability
When we want to see how much is variability with respect to mean.
When score is not same
Eg: English 40/50*100 Compare directly

Math 45/75*100

Standard deviation is the process of making Raw Score into standard Score.

i) Percentage ( % )
Z Score = (X-Xbar)/SD
Z score is a standardized score.

Inferential statistics:

1. Point Estimate ( Decision based on one point )

P value Eg: Marks in exams

2. Internal Estimate
Eg: 95% CI ( CI is confidence interval )
 100-95 time estimate mean value will be under interval
Eg: 30----50--- 70
95 times it will range between 30 and 70.
30 times it will range between 0—30/70—100
Narrow confidence interval is best.
Eg: 90-95% in exam.
95% CI – 95% we are sure that the true population mean would be within the
defined interval range.
Central limit theory:
1. When the size of the individual sample is 30 or more than 30, the sampling
distribution will be equal to true population mean.
Standard error of population is equal to standard deviation devided by square root of
n. ( Number of sample size ) SE =SD/SQRT(n)
2. Frequency distribution of all the sample means is called the sampling distribution.
Sampling distribution is for many means.

Sample distribution is scores.

Normal distribution is also called as bell curve or Gaussian distribution. It can be
mathematically expressed in terms of
Sampling
Distribution of
mean
1. Mean
2. 2. Standard Deviation. Frequency

Mean

The statistical test that we do on data which is not normally distributed is

called non-parametric test.

Variables

Continues Variables Categorical variables or discrete variables

Continues variables are continuous which do not have break which can take
values continuously.
 Discrete variables are without continuity with some error ( distance)
Eg: Gender – Male and Female
Person is live or dead.

Continuous variables Categorical/ Discrete variables

1. Interval 1. Nominal ( name )

2. Ratio 2. Ordinal ( order )

Levels of measurement: Classification of variables.

Level of measurement:

R I O N

1. Continuous or categorical
1. Nominal variables: Where we can give names
Eg: Name : Ram Shyam
Name of the countries, state, types of vegetables

2. Ordinal variables: Anything which can put in an order.

Eg: Ranks – 1st 2nd 3rd
Serial number

3. Interval Variables: Interval is the gap between two points.

Eg: Temperature in form of degree foranheat. In which scales are equally
distributed.

35 36 37 38 39 40 41 42
43 44

Eg: Time. It can never be zero. Ratio is not meaningful. Zero does not have
value.

4. Ratio Variables: Which have all the value of interval variables additional to
it zero value it has and ratio makes meaningful.
 Eg: Money - 1000

Height

Weight

(i) True value of zero ( True absolute zero )

(ii) Ratio is meaningful 50%
(iii) Ratio can be converted into other variables.
Continuous to categorical variables.
Higher measurement Lower measurement
In order to do parametric test, variables should be continuous variables.
When variables are categorical then it is nonparametric test.
Parametric tests have higher power.
1. When data is normally distributed
2. When variance in data is equal
Variance
3. When variables are continuous
4. Variables are Independent
Normally distributed data :
Shapiro Wilk Test
Equality of Variance:
Levence Test

Variables of two groups are equal – Null hypothesis is

No test Continuous in nature -> Level of measurement

Independent  No test.

It is the sin to practice test without using Null Hypothesis

2/11/17
When you choose the design, the type of statistical test you are going to do is almost fixed.
Test for normality ( To check data is bell or not )
Normal distribution – Parametric
Non_parametric - Non_ Parametric
You must know the Null Hypothesis used for all statistical tests.
Equivalance of variance – Lavys Test
Principle of Measurement:
Each & every variable measure should be independent of other measurement.
If assumptions are satisfied parametric tests
If any assumptions are not satisfied non-paramatric tests.