RESEARCH methodology

Dr. S.A.SENTHIL KUMAR

Professor in Management
Pondicherry University
Karaikal Campus
Karaikal – 609 605
Email: [email protected]
 RESEARCH

Research is a scientific and systematic

search for pertinent information on a
specific topic
 RESEARCH PROCESS

Review of Literature
Formulate Design the
Define research Questionnaire
problem Objectives (Determine the Scale)
Find the Research Gap

Interpret and Analyse data Design research

(Test hypothesis if Collect data
report any)
(including sample design)
 Sampling
• The aggregate of all the units pertaining to a study is called the
population. (Total collection of elements about which we study)
• A population is the total collection of elements about which
we wish to make some inferences.
• The listing of all population elements from which the sample
will be drawn is called the sampling frame.
• The process of selecting some of the elements in a population is
called sampling.
• A census is a count of all the elements in a population.
 Sampling Techniques

Elements Selection Probability Non probability

Unrestricted Simple random Convenience

Restricted Systematic Judgment

Stratified Quota

Cluster Snowball
 Probability Sampling
• Simple Random - Each population element has a known
and equal chance of selection.

• Systematic - Every kth element in the population is

sampled, beginning with a random start of an element in the
range of 1 to k

• Stratified - Most populations can be segregated into

several mutually exclusive subpopulations, or strata

• Cluster –The population can also be divided into groups of

elements with some groups randomly selected for study
 COMPARISON OF PROBABILITY SAMPLE
Description Cost and Degree Advantages Disadvantages
of use

Simple Random Sample: High cost Only minimal advance Requires sampling frame to work
Each population element has an Most likely used knowledge of population from; Does not use knowledge of
equal chance of being selected needed; easy to analyse population; larger errors for same
into the sample. data and compute error sample size than with stratified
sampling.

Systematic or fixed interval: Moderate cost Simple to draw sample; If sampling interval is related to a
Researcher uses natural ordering Moderately used easy to check periodic ordering of the
or order of sampling frame, population, may introduce
selects an arbitrary staring point, increased variability.
then selects items at a
preselected intervals.

Cluster sampling: Low cost If clusters geographically Larger error for comparable size
Population is divided into Frequently used defined, yields lowest field than other probability samples.
internally heterogeneous cost; requires listing of all
subgroups. Some are randomly clusters but of individuals
selected for further study. only within clusters

Stratified Random sample: High cost Assures representation of Requires accurate information on
Divides population into mutually Moderately used all groups in sample; proportion in each stratum; If
exclusive subpopulations or strata Reduces variability for stratified lists are not already
and uses simple random on each same sample size available, they can be costly to
stratum. prepare.
 COMPARISON OF NON – PROBABILITY SAMPLE
Description Cost and Degree Advantages Disadvantages
of use

Convenience: Very low cost No need to list of Variability and bias of

Researcher uses most population estimates cannot be
convenient sample or most Extensively Used measured or controlled
economical sample
Purposive: Judgement Moderate cost Useful for certain Bias due to experts’ beliefs
An export or experienced types of forecasting
researcher selects the sample Average use
to fulfill a purpose

Quota Moderate cost Introduces some Bias in researcher’s

Researcher classifies stratification of classification of subjects.
population by pertinent Very extensively population; requires
properties, determines desired used no list of population
proportion of sample from each
class
Snowball: Low cost Useful in locating High bias because sample
Initial respondents are selected members of rare units are not independent.
by probability samples; Used in special populations
additional respondents are situations
obtained by referral from initial
respondents.
 Secondary data
• Data that previously may have been collected
for other purposes but can be used in the
immediate study.
• Examples of secondary sources include the
use of population census data to obtain
information on age, gender, structure of a
population, use of organizations records, use
of hospitals records, and collection of data
from sources such as journals, magazines,
books and other periodicals
 Primary data

• Raw data that has been originally collected

• by the researcher specifically for the study at
hand.
 Methods of Data Collection
Methods of Collecting Primary Data
• Observation Method
• Interview Method
• Collection of Data through Questionnaires
• Collection of Data through Schedules

Difference between Questionnaires and

Schedules
 Analysis of Data

Processing of Data Analysis of Data

Editing Inferential / Statistical

Descriptive and causal
analysis
Analysis

Coding Estimation of
Uni- parameter Testing hypotheses
dimensional values
analysis

Classification
Point estimate Parametric tests
Bivariate
Analysis

Tabulation Non parametric

Interval estimate
Multivariate tests
Analysis
Using
Percentages
 DATA PROCESS

OBJECTIVES OF THE STUDY

DESIGN THE QUESTIONAIRE

DETERMINE THE SCALE OF DATA

FINALIZE THE QUESTIONAIRE

REFER THE OBJECTIVES ONE BY ONE

SELECT APPROPRIATE ANALYSIS

CONSIDER RELIABILITY AND VALIDITY

PRESENT THE RESULT

15
SPSS 16.0 DATA EDITOR WINDOW – DATA VIEW
SPSS 16.0 DATA EDITOR WINDOW – VARIABLE VIEW
 CHARACTERISTICS OF MEASUREMENT SCALES

SCALE NUMBER SYSTEM MARKETING PERMISSIBLE

PHENOMENA STATISTICS
Unique definition Brands Percentages
NOMINAL of numerals Male – Female Mode
(1,2,…………..9) Store types Binomial test
Sale territories Chi-square test

Order of numerals Attitudes Percentiles

ORDINAL (0<1<2………..<9) Preferences Median
Occupation Rank – Order
Social class Correlation
Equality of Attitudes Range
INTERVAL differences Opinions Mean
(2-1 = 7-6) Index numbers Standard deviation
Product – moment
Correlation
Equality of ratios Age Geometric mean
RATIO (0, 2/4 = 4/8) Cost Harmonic mean
Number of Co-efficient of
customers variation
Sales
(Units/Dollars)

18
a parametric statistical test is one that makes assumptions about the parameters (defining properties) of the population distribution(s)
from which one's data are drawn, assume the underlying source population(s) to be normally distributed ,
while a non-parametric test is one that makes no such assumptions.
 OVERVIEW OF UNIVARIATE DATA ANALYSIS PROCEDURES

Univariate Procedures

Interval Nominal
What is the
scale level of
the variable?

Ordinal
1. Descriptive
a) Central A) Mean A) Median A) Mode
tendency B) Standard B) Interquartile B) Relative and
absolute frequency
deviation range by category
b) Dispersion

z test Kolmogorov- Chi-square

2. Inferential t test Smirnov test test

20
 BIVARIATE DATA ANALYSIS PROCEDURES
Bivariate Procedures

Two interval Two nominal

variables variables
What is the
scale level of
the variable?
Two ordinal
variables

1. Descriptive Linear correlation Rank correlation Contingency

coefficient ( ) coefficient coefficient
Simple regression Gamma / Tau Lambda

2. Inferential
* t test on regression
coefficient
Mann-Whitney
* z test on the difference U test Chi-square
between means Kalmogorov - test
* t test on the difference
between means Smirnov test
21
 MULTI VARIATE METHODS

INTERDEPENDENCE DEPENDENCE
METHODS METHODS

1. Factor Analysis Depends on:

2. Cluster Analysis * Number of dependent

variables
3. Multi- dimensional * Nature of the scale of
scaling data

23
Analysis and Interpretation

SAS
FREQUENCY TABLE
 The Chi-Square Test
Chi-square which is available in cross tab is used to test
whether there is a significant association between
two attributes.

Chi-square can be used for three purposes

• To test the normality of distribution
• To test the association
• To test the goodness of fit
 The Chi-Square Test

• Chi-Square Formula :

Expected score:
 HYPOTHESIS:

• Null hypothesis: H0- There is no significant

association between the two variables.

• Alternate hypothesis: H1- There is significant

association between the two variables.
 Result Interpretation
Manual calculation
• CV < TV = Accept the Null Hypothesis (Ho)
• CV > TV = Reject the Null Hypothesis (Ho)

SPSS- Significant Value

• less than 0.05 = Significant = Reject the Ho
• Above 0.05 = Not significant = Accept the Ho
CHI – SQUARE TEST
 ONE WAY-ANOVA
PURPOSE: One way ANOVA is a statistical technique that is used to compare
the means of more than two groups

ASSUMPTION:
• The variables must be normally distributed
• Samples are independent.
• Variances of populations are equal.
• The sample is a simple random sample (SRS).

HYPOTHESIS:
• Null hypothesis H0- There is no significant difference between the
variables.
• Alternate hypothesis H1-There is significant difference between the
variables.
ONE - WAY ANOVA TEST
 PROCEDURE:

STEP 1: Open the database, say INCOME LEVEL.

STEP 2: to test the equality of variance
Analyze-----> compare means--->one-way ANOVA
STEP 3: A one-way ANOVA dialogue box appears as follows
STEP 4:
• select total monthly saving as the dependent variable
• select total income as the factor variable
• Click post hoc ------> Select Duncan.
• Click Continue.
• Click Options----->Select Mean Plot
• Click Continue.
• Click Ok.
STEP 5: Click ok in the one-way ANOVA dialogue box.
 Result Interpretation
ANOVA

Total monthly saving

Sum of Squares df Mean Square F Sig.

Between Groups
3.624E9 4 9.059E8 34.849 .000
Within Groups
1.422E10 547 2.600E7
Total
1.784E10 551

Based on the result generated by SPSS, the significant value is 0.000 and it is
lower than 0.05 so reject null hypothesis . Hence there is a significance
difference in the saving of respondents with respect to their income level.
 Post Hoc: Duncan
Total monthly saving

Duncan
Subset for alpha = 0.05
Total Income N 1 2 3 4
upto 20000 124 5310.48
20001-30000 219 7064.87
30001-40000 103 1.00E4
40001-50000 54 1.12E4
50001 and above 52 1.35E4
Sig. 1.000 1.000 .133 1.000
Means for groups in homogeneous subsets are displayed.

By using Duncan method the saving of respondents is separated into

two groups based on their income level.
 Mean Plot

In the mean plot when the saving level is low when income earned by
them is lower while the saving level is high when the income earned
by them is higher.
 CORRELATION ANALYSIS

• Correlation analysis is a bi-variate technique

• Correlation analysis, refers to the techniques used in measuring the
closeness of relationship between variables of quantitative nature.
• Correlation measures the degree of relationship between variables of
quantitative nature and expresses as a single number
 CORRELATION
PURPOSE: A correlation is a single number that describes the direction and
the degree of relationship between two variables.

ASSUMPTION:
• The variables must be normally distributed.
• The variables must be linear.

HYPOTHESIS:
• Null hypothesis Ho: there is no significant relationship between the
variables.
• Alternate hypothesis H1: there is a significant relationship between the
variables.
 PROCEDURE

• STEP 1: Open the database.

• STEP 2: To run the correlation analysis:
• Analyze------->correlate------------> Bivariate
• STEP 3: The Bivariate correlation dialog box appears
• STEP 4: select the variables
STEP 5: select Pearson correlation coefficients.
• STEP 6: click ok,
CORRELATION ANALYSIS
 TYPES OF CORRELATION
⚫ 1.Simple positive and simple negative : If the values of two
variables move in the same direction, the correlation between
is said to simple positive correlation. If not, then it is said to
be simple negative correlation.
In partial and multiple correlations, three or more
variables are considered simultaneously.
⚫ 2. Partial correlation : In partial correlation, we consider only
two variables to be influencing each other, the effect other
influencing variables being kept constant.
⚫ 3. Multiple correlation : Multiple correlation is defined as a
correlation between one variable and a set of other variables
 SCATTER DIAGRAM METHOD
⚫ It is a simplest method but provides a bird’s eye view
⚫ In this method the given data are plotted on a graph sheet in
the form of dots for each pair of values of X and Y.
⚫ The greater the scatter of dots, the lesser is the relationship
between the variables and vice versa.
⚫ If the dots fall on a st. line from lower left end to upper right
end, they exists perfect positive correlation between the said
variables. On the other hand if line is from the line is from
upper left to lower left, then the relationship is said to be
perfect negative correlation
⚫ If dots are scattered randomly then there exit no relationship
between the variables
TYPES OF CORRELATION – Scatter Diagram
⚫ This method is mainly meant for time dependent data.
The values of the variables are plotted as curves in a
graph sheet.

⚫ If both curves move in the same direction (either

upward or downward), correlation is said to be
positive.

⚫ If both curves move in the opposite direction

correlation is said to negative.
 KARL PEARSON'S CORRELATION COEFFICIENT

⚫ It is a commonly used mathematical method and it qualifies

the degree of relationship between variables

⚫ It is a parametric method

⚫ It is meant for quantitative variables

⚫ This method is to be applied only where the deviations of

items are taken from actual means and not from assumed
means.

⚫ It is defines as r= Coveriance (X,Y) / (Std (X) x Std (Y)

⚫ It is independent of scale and origin

⚫ It can be used for both grouped and ungrouped data

 RANK CORRELATION METHOD
⚫ It is a nonparametric method

⚫ It can be used for quantitative and sometime for

qualitative variables

⚫ It is defined as r=1 (sum square of D) / NXN-N) where

D is different between the ranks of the variables

⚫ If ranks are repeated, adjustment is made by adding

1/12 (mxm-m) to the sum of squae D and m stands for
number of items whose ranks are repeated.

⚫ Pearson and rank correlations will be the same if the

ranks are not repeated
 REGRESSION
PURPOSE: The general purpose of regressions is to learn more about the
relationship between one independent and one dependent variable.

ASSUMPTION:
• The variables must be linear.
• The variables must be normally distributed.

HYPOTHESIS:
Null hypothesis Ho: There is no significant relationship between the variables.

Alternate hypothesis H1: There is a significant relationship between the

variables.
 PROCEDURE:
• STEP 1: Open the database, say Family Income and Total
Monthly savings.
• STEP 2: To run a linear regression analysis
• Analyze → Regression → Linear.
• STEP 3: Linear regression dialog box appears as below.
• Select Total monthly saving as a dependent variable
• Select Total income as a independent variable
• Click plots
• Select histogram and normal probability plot
• Clicks Continue.
• Click save in the linear regression dialogue box
• Select standardize in the predicted value group
• Click Continue.
 Model Summaryb

Model R R Square Adjusted R Square Std. Error of the Estimate

1
.404a .163 .162 5210.879

a. Predictors: (Constant), Total income

b. Dependent Variable: Total monthly saving

 MULTIPLE REGRESSION
PURPOSE: The general purpose of regressions is
to learn more about the relationship between
one dependent and more than one independent
variable.

ASSUMPTION:
◼ The variables must be linear (Metric) – Ratio or

Interval scale.
◼ The variables must be normally distributed.

53
 FACTOR ANALYSIS
◼ Factor analysis is a technique that is used
to reduce a large number of variables into
fewer numbers of factors.
◼ Factor analysis extracts maximum common
variance from all variables and puts them
into a common score.
◼ Interval data are assumed for factor
analysis.
54
 Cluster Analysis

◼ Cluster analysis can be defined as a set of

techniques used to classify the objects into
relatively homogeneous groups called clusters
◼ It involves identifying similar objects and grouping
them under homogeneous groups
◼ Cluster as a group of objects that display high
correlation with each other and low correlation
with other variables in other clusters
Thank you

56