Introduction to Biostatistics

Dr. Shakil Shams

M.B.B.S., M.Phil. (Anatomy)
Assistant Professor
Department of Anatomy
Dhaka Medical College, Dhaka
 Topics outline

 Statistics and Biostatistics

 Research

 Research methodology
 Data

 Variable

 Scales of measurement
 Study design
 Sampling & sampling methods
 Data presentation
 Topics outline (Contd.)

 Probability, P-value & probability distribution

 Data interpretation
 Data analysis
 Estimation

 Hypothesis & hypothesis testing

 Tests of Significance
 Sensitivity & specificity
 Correlation & regression
STATISTICS & BIOSTATISTICS
 Statistics

It is the science which deals with development and

application of the most appropriate methods for the –
Collection of data

• collection of data
• presentation of collected data
• analysis & interpretation of the results
• making decisions on the basis of such analysis
Statisticians try to interpret and
communicate the results to others
 Biostatistics

It is the branch of applied statistics directed toward

applications in the health sciences and biology.
 Why biostatistics?
 some statistical methods are more heavily used in
health applications than elsewhere e.g. survival
analysis

 examples are drawn from health sciences - makes

subject more appealing to those interested in health

 illustrates how to apply methodology to similar

problems encountered in real life health situations
 Biostatistics (Contd.)

Statistics is not merely a compilation of

computational techniques- it is a way of learning
from data

Biostatistics is concerned with learning from

biological, public health, and other health data

Basic Biostatistics 9
Biostatisticians are:
Data detectives who uncover
patterns and clues through data
description and exploration

Data judges who confirm and

adjudicate decision using
inferential methods

Basic Biostatistics 10
Basic Biostatistics
 Knife for Surgeon
Biostatistics for Medical researcher
RESEARCH
 Research

Systemic, scientific and ethical search to

explore new knowledge for solving a particular
problem
 "Exitus" (death) table compiled by Dr. Sigmund Rascher

Body
Body
Water temperature
Attempt no. temperature at Time in water Time of death
temperature when removed
death
from the water

5.2 °C 27.7 °C 27.7 °C

5 66' 66'
(41.4 °F) (81.9 °F) (81.9 °F)
6 °C 29.2 °C 29.2 °C
13 80' 87'
(43 °F) (84.6 °F) (84.6 °F)
4 °C 27.8 °C 27.5 °C
14 95' 95'
(39 °F) (82.0 °F) (81.5 °F)
4 °C 28.7 °C 26 °C
16 60' 74'
(39 °F) (83.7 °F) (79 °F)
4.5 °C 27.8 °C 25.7 °C
23 57' 65'
(40.1 °F) (82.0 °F) (78.3 °F)
4.6 °C 27.8 °C 26.6 °C
25 51' 65'
(40.3 °F) (82.0 °F) (79.9 °F)
4.2 °C 26.7 °C 25.9 °C
29 53' 53'
(39.6 °F) (80.1 °F) (78.6 °F)
Is this research?
Is this research too?
“Among all criminals and murderers, the most

dangerous type is the criminal physician.” 

Miklos Nyiszli, prisoner at Auschwitz

Nuremberg code of Medical Research,1947:

 adopted 10 principles

 informed written consent from the participant is

the first pre-requisite

 Purpose of research

 discover new facts or principles

 verify and test old facts or principles

 fresh interpretation of known fact

 develop tool, concept, theory

 Types of research

 Basic research
 Applied research
 Quantitative research
 Qualitative research
 Descriptive research
 Analytical research
Basic (pure) research
• deals with basic processes of life, disease & social /
natural phenomenon

• generate new ideas, principles & theories

• intellectually & academically interesting

• has no immediate utility for any existing pressing

problem

e.g.
 How does a malignant cell multiply?

 How genes are regulated?

Applied (practical) research

• existing problem oriented research directed to solve

an immediate pressing problem

• inspired by the need of social action

e.g.

risk factors of early onset MI

prevention of high infant mortality

Quantitative research
• deals with phenomenon that can be quantified and

expressed in terms of quantity/amount

• measures and compares the magnitude of a problem

or phenomenon

• data collected in numbers

e.g.
height, weight, skinfold thickness, upper arm and calf
girth, breadth of humerus and femur
study on dietary program for reducing body weight
Qualitative research
• deals with the phenomenon that can not be quantified

and can not be expressed in terms of quantity

• explains, describes and understands the phenomenon

to answer about the complex nature of it

• data collected in words

e.g.
 People’s opinion about the services provided by
Dhaka Medical College Hospital

 Perception of security at work place by the female

garment-manufacturing workers
Descriptive research

• describes situations and events as it exists/ed naturally

to answer the questions who, what, when, which and

where

• researcher has only to report what is happening or

what has happened - no explanation & no cause-effect

relation

e.g.
prevalence, distribution & pattern of goiter

antenatal care practice in rural area

Analytical (explanatory) research

•explains the reasons of the phenomenon that the descriptive

research observed

•deals with the determinants of the phenomenon

•attempts to establish Cause-Effect relationship between

variables

•use facts already available & analyze these to make critical

evaluation
 Analytical research (Contd.)

•answers to the question how/why

e.g.
Why goiter is more common in some areas of Bangladesh?
Is analytical research superior to
descriptive research?
 Stages of research
1) Planning of research

• Generation of idea and problem identification

• Knowledge building about the problem (review of literature)
• Statement of the research problem
• Statement of research question (RQ)
• Statement of research hypothesis
• Setting of research objective
• Deciding on appropriate study population & study design
• Deciding on appropriate sample size & sampling technique
• Deciding on data collection plan
• Development of data collection instrument
 Stages of research (Contd.)

2) Implementation or data collection

3) Data management
• Data editing
• Data reduction
• Data presentation
• Data analysis
• Data interpretation
• Data inference / decision

4) Report writing

5) Dissemination / publication of the report

 Some of these steps may go
simultaneously & repeatedly
throughout the research process
List of topics :
 Research methodology & research method
 Some steps of planning of research
 Research problem
 Knowledge building/ literature review
 Research title
 Research question(s)
 Research hypothesis
 Research objective(s)
 Research questionnaire
Research methodology
“The grand aim of all science is to cover the greatest

number of empirical facts by logical deduction from

the smallest number of hypotheses or axioms.”

Albert Einstein
Research methodology

 It is the way to deal with the various steps adopted by a

researcher to study the research problem systematically
along with the logic, assumptions and rationale behind
them.

 Whenever we choose a research method, we must justify

why we preferred this over others. Research methodology
seeks to answer this question.
The two terms research methodology and
research method are often used
interchangeably. This is incorrect.
Research methods

 Techniques and tools used for conducting research.

 These are -

1. methods dealing with collecting & describing data

2. techniques used for establishing statistical relationship

between variables e.g. statistical tests, correlation-regression
analysis, odds ratio etc.

3. methods used to evaluate the reliability, validity and

accuracy of the results delivered from the data e.g.
sensitivity, specificity, PPV, NPV etc.
 When we speak about research methodology we not
only talk about the research methods but also keep in
view the logic behind the method that we use in the
context of our research undertaken.

 Research methodology has many dimensions; research

methods are part of research methodology.
Research problem
 Research problem

 Research problem is a perceived difficulty or a feeling of

discomfort which a researcher experiences due to the

discrepancy between the existing situation and what it

should be
“যা চেয়েছি কেন তা পাই না?”

Research problem

“যা পেয়েছি কেন তা চাই না?”

Research problem
Problem = Expectation – Reality
 How does research problem arise?

Research problem arises from:

 Day-to-day personal experiences

 Practical issues in the hospital and community

 Findings of the previous researches

 Brain storming

 Intuition
 Criteria of research problem

 Must be -

 researchable

 important

 feasible

 ethically acceptable

 Should be interesting
Knowledge building
 Knowledge building
 for obtaining in-depth knowledge & insights about the research
problem
 done by:

 literature review
 review of books, conference proceedings, reports
 thesis, newspaper, magazines
 experience survey
 focus group interview
 case study
 pilot study
Purpose of Knowledge Building:
 avoidance of duplication / repetition
 finding of gaps / conflicting information
 identification of unanswered questions
 discovery of fallacy / inconsistency
 identification of study variables
 development of research question, hypothesis, objective

 idea generation about population, study design, sample

size, sampling, statistical process & study procedure
 Knowledge building about the problem

 Before undertaking research work 60-70%

 During the research work 10-20%

 After completing data analysis but

15-20%
before writing the research report

These proportions are not rigid.

Research title
 Research title

 Should be accurate, complete and specific

 For accuracy, should use the same terms in the title as in

the question and answer

 For completeness, should include all information that

reflect all the main topics

 For specificity, should use specific words

 Research title (Contd.)
 Title should be unambiguous

 avoid noun clusters

 avoid misplaced adjectives
 not use abbreviations

 Title should be concise

 <100 characters and spaces

 omit unnecessary words
 omit nonspecific openings such as ̎Studies of ̏
 omit ̎ the ̏ at the beginning of the title
Research question
 Research question

 Queries or ideas arising out of the research problem for the

researcher seeks answer through his or her research effort

 research question is a question, not a statement

 there may be more than one research question in a research

e.g. Does Hb. concentration fall in pregnancy?

 Types of research question

 What

 Where

 When Descriptive study

 Who

 Which

 How
 Why Analytical study
 What is the effect of exercise on serum cholesterol?

 How different foot dimension are related to stature?

 Who are affected in machine handling in garments factory

regarding hand grip strength?

 Where is the drug found in highest concentration after its

intravenous administration?

 Why 2D:4D is sexually dimorphic amongst adult

Bangladeshies?
 Criteria of a good research question

FINER
 feasible

 interesting

 novel

 ethical

 relevant
 How to develop a good research question:

1. begin by identifying a broader subject of interest that lends

itself to investigation e.g. childhood obesity

2. do preliminary research on the general topic to find out what

research has already been done and what literature already
exists

3. find a unique area that yet to be investigated or a particular

question that may be worth replicating
 How to develop a good research question (Contd.)

4. begin to narrow the topic by asking open-ended "how" and

"why" questions e.g. consider the factors that are
contributing to childhood obesity or the success rate of
intervention programs

5. create a list of potential questions for consideration and

choose one that interests you and provides an opportunity for
exploration
6. Finally, evaluate the question by using the following list of
guidelines:

 Is the research question interesting? 

 Is it a new issue or problem that needs to be solved?
 Is it attempting to shed light on previously researched topic?
 Is the research question researchable?  
 Is the methodology to conduct the research feasible?
 Is the research question measureable?
 Will the process produce data that can be supported or
contradicted? 
 Is the research question too broad or too narrow?
 Too narrow Better

What is the childhood obesity rate in  How does the education level of the
Dhaka?    parents impact childhood obesity rates
in Dhaka?

This is too narrow because it can be This question demonstrates the correct

answered with a simple statistic.  amount of specificity and the results

would provide the opportunity for an

Questions that can be answered with a
argument to be formed. 
"yes" or a "no" should typically be

avoided.
 Unfocused and too broad More focused

What are the effects of childhood How does childhood obesity correlate with
obesity in the Bangladesh?  academic performance in elementary
   school children?

This question is so broad that research

This question has a very clear focus for which
methodology would be very difficult.
data can be collected, analyzed, and discussed.

The question is too broad to be discussed in

a typical research paper.

 Too objective More Subjective

How much time do young children What is the relationship between

spend doing physical activity per day?  physical activity levels and childhood
   obesity?

This question may allow the researcher to This is a more subjective question that may

collect data but does not lend itself to lead to the formation of an argument based on

collecting data that can be used to create a the results and analysis of the data.

valid argument because the data is just factual

information.
 Too simple Better

How are school systems addressing What are the effects of intervention
childhood obesity?  programs in the elementary schools on
the rate of childhood obesity among 3rd -
6th grade students? 

This information can be obtained without the This question requires both investigation and

need to collect unique data. The question could evaluation which will lead the research to form

be answered with a simple online search and an argument that may be discussed.

does not provide an opportunity for analysis.

PHRASING OF RESEARCH QUESTION

DICTATES TO THE APPROPRIATE STUDY

DESIGN
 Research question Study design needed

How severity of DM is related with food habit? Cross sectional study

Is high serum uric acid concentration in eclampsia

Case control study
associated with adverse fetal outcome?

What is the effect of exercise on serum cholesterol?

Cohort study

Is there any influence of spirulina on experimentally

Experimental study
induced atherosclerosis in rats?

Is laparoscopic cholecystectomy better than traditional

Clinical trial
surgery?
Research hypothesis
 Research hypothesis
 It is a logical, tentative(assumed) & testable answer(s) to the research
question

 Concerned with the parameters of the population about which the

statement is made

e.g.

Hemoglobin concentration falls during pregnancy.

Hypothesis reflects the depth of knowledge and the capability of

imagination of the researcher
Importance of hypothesis:

 Help to select methodology and methods of data collection

 Help to select population and variables to be studied

 Help to select intervention needed

 Types of hypothesis

1. Null hypothesis (Ho) :

It is the hypothesis of no difference.

It states –

 no real difference between sample statistics &

population parameter

 observed result is purely due to by chance or sampling

error
 
Examples:
 
(1) Mean cholesterol value in normal (M1) = Mean cholesterol

value in hypertension patients ( M2 )

 
(2) No association between lung cancer and smoking
2. Alternative hypothesis (Ha):

It is the hypothesis of difference.

It states –

 sample statistic is different from population parameter

 observed result is not due to by chance or sampling error

rather due to some valid reason or extraneous factor (real

difference)
Examples:
 
(1) Mean cholesterol value in normal (M1) <

mean cholesterol value in hypertension patients ( M2 )

 

(2) There is association between lung cancer and smoking

* In fact, alternative hypothesis is the researcher’s

hypothesis
H0 & Ha are diagonally opposite
• Research question is must for all research.

• Research hypothesis is needed in

analytical research.
Examples of good and bad hypothesis statements
 Steps of constructing a good hypothesis

1. State the research question:

“How does a person’s level of education influence their

attitudes towards immigrant rights?”

 Steps of constructing a hypothesis (Contd.)

2. Write a theoretical statement explaining why you think the

independent variable will increase or decrease the dependent
variable.

Possible ways education could affect immigration attitudes:

 People with higher levels of education are more politically liberal
than those with low levels of education, so people with higher
levels of education will be more supportive of immigrant rights.
 People with less education are more likely than people with more
education to know someone who has recently immigrated to the
U.S., so people with lower levels of education will be more
supportive of immigrant rights.
 Steps of constructing a hypothesis (Contd.)

3. State the alternative hypothesis.

“People with a college degree will agree more strongly than

those with no college degree that legal immigrants should
have the same rights as U.S. citizens.”

 
 Steps of constructing a hypothesis (Contd.)

A hypothesis must be falsifiable. This means the hypothesis

can be proved wrong.

So,

4. Finally state the null hypothesis.

“People with a college degree do not agree more strongly than

those with no college degree that legal immigrants should
have the same rights as U.S. citizens.”
Some points to note :

Hypothesis -

 gives only the tentative explanation of the research question

 is a statement , not a question

 is not a must for a research

 may or may not be the real situation

is to test, not to prove

It is better for the researcher to follow the research
questions, instead of formulating hypothesis for
the problems on which no research has been
carried out so far
Research objective
 Research objectives (SMAART)
• Objectives are goals to be achieved through the research
process
• Reflects the questions whose answers the researcher wants
the study to yield
• Objective covers –
 exploration
 description
 explanation
 prediction
 evaluation
 impact assessment
 Research objectives (Contd.)

 General objective:

 overall goal /aim of the research process

 short statement about what is expected to be

achieved by the research
 Research objectives (Contd.)

 Specific objective:

every individual task or work that is to be done to

achieve the general objective

 Ultimate objective:

statement that tells about the benefits, implications and

utilization of study findings
 Example (early onset MI)

Research problem:
• Now-a-days, people are dying from MI in very early age.

Research question:
• What are the risk factors of early onset MI?
Research hypothesis:

• Obesity, HTN, DM & dyslipidemia are the causes of early

onset MI.

General objective:
• to find out the risk factors of early onset MI
Specific objective:
• to measure height & body weight of study subjects
• to measure blood pressure of study subjects
• to measure fasting blood glucose of study subjects
• to measure lipid profile of study subjects

Ultimate objective:
• to make the people aware about the risk factors of early
onset MI to reduce the burden of early onset MI
Research questionnaire
 Research questionnaire

A questionnaire is a data collection tool containing series of

questions that is generally mailed or handed over to the

respondent and filled in by the respondent him/herself or by

the interviewer in favour of the respondent.

Questionnaire should be be simple & clear
Questionnaire should also be adequate in length
“A good speech should be like a woman's skirt; long
enough to cover the subject and short enough to
create interest.”
Winston S. Churchill
Research is not …
an accidental discovery

accidental discovery may occur in structured research

process

merely collection of data

collecting reliable data is part of the research process

searching out published research results in libraries /

internet

research process always includes synthesis and analysis

but just reviewing of literature is not research
Research is a creative and circular process
A brave researcher sits behind the target to see what he can see
&
never distorts the facts even it is found contra to be

Basic Biostatistics 99
DATA & VARIABLE
 “Data! data! data!” [Holmes] cried impatiently. “I can't make

bricks without clay.’’

DATA:

a set of values recorded on observational units

VARIABLE:

characteristic or attribute of an individual / object /

phenomenon that take on different values in different

persons / objects or in the same person / object in different

time, place etc.
Example:

Variable Data

Blood pressure 120 mm of Hg, Hypertension

Age 50 yrs., Old, Young

 Types of data
1.Qualitative (categorical) data

• data that vary in kinds

• expressed as rate, ratio, percentage, proportion
• have no scale of measurement

• provides answer to the question : What type?

• two types :
a) Nominal : categories cannot be ordered one
above another e.g. Sex , Marital status

b) Ordinal : categories can be ordered one

above another
e.g. Level of knowledge, Pain score
2.Quantitative (numerical) data
• data that vary in amount and can be measured and ordered in
terms of quantity
• expressed as mean, range etc.
• have scale of measurement

• provides answer to the question : How much?

• two types –
a) Continuous : take any value even fractions or decimals
e.g. height, weight

b) Discrete : take only whole numbers

e.g. family member
Exercise 1:
To check the accuracy of the clinical diagnosis of malaria,
blood slides from 33 patients were examined for MPs.
There were three possible results : Negative, P.
falciparum or P. vivax.

The results were :

Negative 19
P. falciparum 13
P. vivax 1
TOTAL 33

These data are:

Nominal/Ordinal/Continuous/Discrete
Exercise 2:
Health personnel from 148 rural health institutions were
asked the following question : “How often have you run
out of anti-malarial drugs in the last two years?”
There were four possible answers : never , 1 to 2 times
(rarely) , 3 to 5 times (occasionally) , more than 5 times
(frequently).
The results were :
Never 47
Rarely 71
Occasionally 24
Frequently 6
TOTAL 148
These data are:
Ordinal
At the end, statistics is a game of numbers, be it a
qualitative or quantitative data.
 Types of variable

1. Independent variable (Usually the cause)

• variable that influences, regulates or cause some changes

in the dependent variable

• selected for the study in the believe that it is a contributory

factor or at least can influence the problem

• has causal or input or exposure status

2. Dependent variable (Usually the effect)

• a measure that reflects the effect of independent

variable

• selected for the study in the believe that it helped to

describe the problem

• has output or outcome or effect or response status

DOORE
Example :

Independent V. Dependent V.

Salt intake Hypertension

Hypertension MI

Finger tip to root Digit 3 Hand grip strength

 3. Intervening (Intermediate) variable:

• 3rd variable through which independent variable affects the

dependent variable
• fits into a causal chain
e.g.

Independent V. Intervening V. Dependent V.

Salt intake Hypertension MI

Low economic Inadequate diet Underweight

status
Example from research setting :

It is expected that the incidence of diarrhoea would

decrease as the number of water faucets in a village

increased. If there is no change over time, there might

be an intervening variable.

People, for example, may dislike the taste of tap-water so

much that they use it for everything, except for

drinking.
4. Confounding (Extraneous) variable:

• 3rd variable that is independently related to both

dependent & independent variables and thereby might
affect the relationship between these two
• not in a causal chain
• not related to the purpose of study
• distort the study result
e.g.
Obesity MI

Hypertension
Possible combinations:

Obesity (+) Hypertension (+) MI (+)

Confounding variable
Obesity (+) Hypertension (+) MI (-)

Obesity (+) Hypertension (-) MI (+)

Obesity (+) Hypertension (-) MI (-)

Obesity (-) Hypertension (+) MI (+)

Obesity (-) Hypertension (+) MI (-)

Obesity (-) Hypertension (-) MI (+)

Obesity (-) Hypertension (-) MI (-)

 Other terms related to data

Primary data
• Obtained 1st hand by researcher
• Generated by observation, measurement, interview etc.

Secondary data
• Obtained by some others
• Have already passed through statistical processes
• Collected from records, documents, journals , other studies
Derived data
• Derived from primary or secondary data

e.g. BMI derived (calculated) from body weight and height

Dichotomous (Binary) data

• Expressing only two mutually exclusive information

e.g. Sex (male or female)

Univariate data
• Express only one information

e.g. Birth rate of female baby

Bivariate data
• Express two linked/related information simultaneously

e.g. Birth rate of Rh+ve female baby

Multivariate data
• Express more than two related information simultaneously
e.g. Birth rate of Rh+ve female premature baby

Outlier data
• Distinct from the main body of data
• Incompatible with the rest of the data
• Usually regarded as error but may be true also
SCALES OF MEASUREMENT
 Scales of Measurement

Nominal: Classification

Ordinal: Ranking

Interval: Equal interval

Ratio: Absolute zero

Scale Classification Order Equal Intervals Zero

Nominal Yes No No No
Ordinal Yes Yes No No
Interval Yes Yes Yes No
Ratio Yes Yes Yes Yes
 Scales of measurement (Contd.)

Nominal scale:
• Assigns some numerical values to the

variable only for-

identification

e.g. Student registration no. 1,2,3 …

 Scales of measurement (Contd.)

Ordinal scale:
• Assigns some numerical values to the

variable for-

identification

ranking

e.g.

Rich - 1

Middle class - 2

Poor - 3
 Scales of measurement (Contd.)

Interval scale:
 Contains highest & lowest values
 Intervals between adjacent scale values

are equal

 Has an arbitrary zero

 Measures on either side of zero

e.g. Celsius & Fahrenheit

temperature scales, IQ scores
 Interval scale (Contd.)

For –

identification

ranking

addition

subtraction
 Ratio scale:
 Contains highest & lowest values
 Intervals between adjacent scale values are

equal
 Has an absolute zero (zero means zero)
 Measures on only one side of zero

e.g. Kelvin temperature scale, income, length,

area, volume, height, weight

 Ratio scale (Contd.)

For–

identification

ranking

addition

subtraction

multiplication

division

Highest scale - can be used for any purpose

 20

15

10

-5

-10

-15

-20
Interval scale Ratio scale
 Hint:
If score can go below zero or if
the zero is not a true zero, measurement is
interval.

If score can not go below zero or

if the zero is a true zero, measurement is ratio.

R AB
Possible data types and levels of measurement
Level of measurement is set for each variable to
obtain the required statistical test &
presentation in SPSS
STUDY DESIGN
 STUDY DESIGN

 Scientific and ethical methods of search to collect valid

and reliable information / data

Features of study design:

 ethical

 capable to obtain reliable and valid data

 help researcher to avoid wrong conclusion
 Selection of a study design depends on -

 type of research problem

 resources available

 time to complete the research

 knowledge about the research problem

Some core concepts
Validity / accuracy: degree of closeness of a measurement or
a test to the fact (true value)

Reliability / precision: quality of a test to get the consistent

results over time when repeated in identical situations
 Example : BP of an individual is 120 mmHg.

The center of the target represents the true value of the

substance being tested (Here, 120mm of Hg).

Reliable Reliable &

not valid valid

Not reliable Data set-2

Data set-1 not valid
119, 120, 122

88, 89, 90 Data set-3

88, 110, 150

Temporal relationship:

Time sequence between exposure /input/cause and

outcome/result/effect

 Exposure always precede the outcome

 Outcome always follows the exposure

“ ডিম আগে না মুরগী আগে ??? ”

Exposure Outcome

Obesity MI

Smoking Lung cancer

 Prospective study
 Measurement of exposure prior to the occurrence of
outcome
or

Data recorded or generated about the events that

will occur after the start of the study

 Forward looking study

 Moves from exposure to outcome

 Retrospective study
 Measurement of exposure after the occurrence of
outcome
or

Data recorded about the events occurred in the past

before the start of the study

 Backward looking study

 Moves from outcome to exposure

Prospective study:

Exposure Outcome

Retrospective study:

Outcome Exposure
•Spatial relationship:

Related to space
e.g.
Change in spatial
configuration of
troponin during
muscle contraction

“ আমি কোথায় ??? ”

 Longitudinal study
• Data collection at more than one point of time with

follow up on same study subjects

•Usually prospective but may be retrospective as well

e.g. Clinical outcome of CABG in CAD

Rate of RTA of a city in last 10 years

Onset

Longitudinal & Longitudinal &

Retrospective Prospective
 Cross sectional study

Single time data collection on a cross section of

population at one definite point of time or within a
short span of time
 No follow up & no repeated data collection

e.g. Recording the ECG findings of an MI patient at the

time of admission – just for once
Types of study design
 Types of study design
Observational (non-interventional):
a. Descriptive
1. Case study
2. Cross sectional study
3. Surveillance
b. Analytical
1. Case control study
2. Cohort study
3. Cross sectional study
Experimental (interventional):
1. Clinical trial
2. Community intervention trial
3. Quasi experimental study
OBSERVATIONAL vs. EXPERIMENTAL STUDY

Outcome

Cause/exposure Outcome
 Observational / Non-interventional
study

 Based on observation of events created either by nature

(by natural phenomenon) or by human beings (but not for
research purpose)

 Researcher does not manipulate the outcome

A sample of population is observed for different

characteristics by interview, questionnaire, measurement,
records etc.
 Study on “Correlation of stature with measurements of
foot segments calculated from footprint and foot
outline of adult male Bangladeshis”

 Study on “Health hazards after an earthquake”

e.g. “Hand anthropometry and hand grip strength of adult
female Bangladeshi garment workers”
Types of observational study
 Types of descriptive study

a) case study/case report - for rare diseases or common diseas

with uncommon presentation

CA R D or CA RP

b) case series- repeated case study

c) cross sectional study- when comparison and follow-up

is not done
 Analytical study
 Synonyms - Comparative study

Explanatory study

Causal study
 Needs a comparison group to come to a conclusion
 Explains disease occurrence
 Analyzesand answers specific research question/s
 Research hypothesis is a must
 Analytical study (Contd.)

 Primary goal is to establish the association between the

exposure/ risk factor/etiology and

outcome/disease, so provides etiology or
determinants of the problem

 Types are:
a) case control study
b) cohort study
c) cross sectional study-when comparison is done
 Descriptive vs. Analytical Study

1. Describes the distribution 1. Describes the causes

of problem or determinants of
problem
2. Needs no comparison 2. Comparison group
group needed
3. No attempt to analyze 3. Cause-effect
cause-effect relationship relationship analyzed

4. Usually no hypothesis 4. Hypothesis testing is

testing is done done
 Distinction between descriptive study
& analytical study is not so clear cut
A large scale descriptive study may give a clear
answer to specific research question
An analytical study may be incidentally of great
descriptive interest
 Case control / Case referent study
 Moves from outcome to exposure
Outcome-based sampling is done. Study subjects are
selected based on the outcome (present/absent)
Cases are with outcome and controls are without outcome
 Outcome not always has to be disease
Mostly retrospective and longitudinal
 Case-control ratio - 1:1 or maximally up to 1:4
 Types of case control study

 Population based : case and control from the same base

population
 Hospital based : case and control are selected from
hospital admitted patients
 Multi-factorial : several exposures are explored
simultaneously
 Nested : case and control are selected from the same
cohort and nested within the cohort
 Advantages of case control study

 Quick, inexpensive
 Small sample size
 Good for rare and chronic disease
 Less ethical constrain

CA R D
 Disadvantages of case control study

 Can not infer temporality between exposure and

outcome
 Information on exposure may be less accurate
 Prone to bias
 Cohort study
 Cohort: Special group of people having some definite, common
base line characteristics and exposed to the same environment
for a long period and who are followed up for a definite period

e.g. radiologists, garment workers, army personnel, mine

workers etc.

CO R E
 Good for rare exposure

e.g. Alcohol consumption in antenatal period

 Cohort study (Contd.)

 Moves from exposure to outcome

 Exposure-based sampling is done. Study subjects are
selected based on the exposure(present/absent)
 Exposed group (with exposure) & unexposed group
(without exposure)
 All must be free from outcome at the start of study
 Mostly prospective and longitudinal
 Cohort study (Contd.)

e.g.
a) Study on prevalence of respiratory tract infections of
individuals working in anatomy dissection hall

b) Study on the association of deposition of metals in the

lungs of coal miners
 Cohort study (Contd.)

 Researcher identifies a cohort population with and

without the exposure status but all must be free of
outcome at the start
 follows the cohort in future with observation at
several points of time and determines the outcome in
exposed and unexposed group
Design of cohort study
 Types of Cohort Study

1. Prospective cohort study

2. Retrospective or Historical cohort study

 Exposed (CO) Poor SAP

Good SAP
Cohort

Poor SAP

Unexposed (no CO)

Good SAP

Time
Onset
Direction of inquire
Prospective cohort study on childhood obesity
(CO) & school academic performance (SAP)
 Historical cohort study
In 2016, a historical cohort study can be designed to
study the effect of neonatal asphyxia on future
neurological disability by conceptually going back to
1996.
 Retrospective as well as prospective
 Historical cohort study (Contd.)

• Outcome already occurred before study onset.

• Exposure base in past

• Cohort is defined in past based on previous dat.

• Follow up directed (through records) from past to

present up to certain cut-off time
 Exposed (CO) Poor SAP

Good SAP
Cohort

Poor SAP

Unexposed (no CO)

Good SAP

Time
Onset
Direction of inquiry
Historical cohort study on childhood obesity (CO)
& school academic performance (SAP)
 Prognostic Cohort Study

Identify factors influencing the prognosis of disease after

diagnosis & treatment.

Cohort composed of cases diagnosed and treated & then

they are followed up to evaluate prognosis with respect
to some factors.

Here cohort cases are not free from disease but free from
outcome of interest (cure, death, disability etc).
 Death
MI with DM

Cured
MI

Death

MI without DM
Cured

Time
Onset
Direction of inquiry

Prognostic cohort study of DM on MI

 Advantages of cohort study

 Possible to measure multiple outcomes against a

single exposure
 Less prone to bias
 Can ensure temporality
 Good for rare exposure
 Measures incidence
A wide picture of health hazard can be obtained
 Disadvantages of cohort study

 Costly, time consuming

 Large sample size needed
 Chance of attrition- lost to follow up due to
withdrawal, death, change of location
 Not good for rare disease
 Experimental / interventional study

 Based on experiment, created by the researche

 Researcher determines who will be exposed to
the factor of interest and who will not
 Researcher intervenes to affect the outcome
 More ethical and feasibility issues
 Less chance of bias
e.g. To reduce the BP
 give drug
 change of lifestyle
Intervention may be:
 Drug therapy
 Treatment regime
 Surgery

 Dietary manipulation
 Change of lifestyle
 Medical counseling
 Rehabilitation procedure
 Blinding or masking
 Ignorance of certain person/s involved with the
clinical trial - regarding the intervention assigned to
the participants

Advantages:
 Avoid assessment bias
 Protect the behavioral change of the participants
Types :
1. Single blind: participants are kept ignorant

2. Double blind: participants and assessors

are kept ignorant

(Common type)
3. Triple blind: participants , assessors and

researchers - all are ignorant

e.g. Experiments at Guantanamo Bay prison

 Blinding (masking)

Blinded
Type
Participants Assessor Researcher

Single blind Yes No No

Double blind Yes Yes No

Triple blind Yes Yes Yes

Experimental study:
1.Clinical trial
2.Community intervention trial
3.Quasi experimental study
 Clinical trial
 It is a prospective and experimental study comparing

the effects of intervention in human subjects

Types:

a) controlled clinical trial

b) uncontrolled clinical trial

c) randomized controlled clinical trial

 Controlled clinical trial
 It is a clinical trial comparing the effects of
intervention in an experimental group against a
control group involving human subjects to
determine which of the intervention is of greatest
benefit
 Placebo
 Preparation identical in all respects to that given to the
treatment group except that it lacks the active
component
 Look and taste must be equivalent to the active drug
used
 Conditions to allow placebo as ethical:

 No standard treatment is available

 Existing treatment has doubtful efficacy
 Existing treatment is rarely available to population at large
 Patient is not benefited by standard treatment and there is no
second option
 Patient refuses existing treatment and willing to be on
placebo
 Patient suffers from minor disease
 Test regime is an add on to existing regime
 Types of control group (Contd.)

 Concurrent control: independent control group is

generated along with intervention group who receive
placebo or equivalent intervention for the same period as
experimental group

 Non-concurrent / historical / external control: results of

study by another researcher or same researcher done in the
past on an identical issue is used as control
 Uncontrolled clinical trial
 Clinical trial which evaluates the effects of
intervention in an interventional group of human
subjects without comparison with any control group

 Here, control group design is not ethical or possible

e.g.

Ethically, placebo should not be given to a Ca-

breast patient. So, Ca-breast patient can not be a
control.
 Randomized controlled clinical trial
 Clinical trial comparing the effects of intervention in
an experimental group against a control group

following random allocation of participants to

interventional and control group

 Randomization: random allocation of study subjects

in experimental or control groups by lottery technique
 Features of randomized clinical trial
 Interventional: intervention is done to manipulate the
outcome

 Controlled: control group is present

 Randomized: participants are allocated randomly in

interventional and control groups

 Prospective, longitudinal, analytical type of study

 Stages of randomized clinical trial

a) Enrollment: by randomized sampling followed by

inclusion and exclusion criteria

b) Allocation: by randomization into experimental and

control groups

c) Intervention: given to both groups

d) Follow up: to observe the outcome in both groups

e) Analysis: comparison of outcomes

Non-random
sampling
 Inclusion criteria
 Equivalent to diagnostic criteria
 Strict criteria to identify target group for clinical trial
 Avoids selection bias & misclassification of participants

e.g. For a study to be done on “Different dimensions of

foot of adult male Bangladeshis”- the inclusion criteria
will be,

a) 20-25 years of age (age confirmed by

national ID)
 Exclusion criteria

 Takes into account:

 Confounding variable/s

 Co-morbidities

 Non-compliance

 Refusal to participate
 Exclusion criteria (Contd.)

e.g. For a study to be done on “Different dimensions of

foot of adult male Bangladeshis”- the exclusion
criteria will be,

a) known case of congenital or acquired

foot deformity

b) history of trauma to foot

 Parallel design / Concurrent controlled CT
 Participants are selected by RCT. Then, intervention will
be given to the experimental group & placebo will be
given to the control group.
 Widely practiced.
Cross-over design / Concurrent & self-controlled
CT

 If question arises that the effect in the experimental

group may not solely be due to the intervention rather
due to some other factors – then cross-over design is
employed.
Cross-over design (Contd.)
 Community interventional trial
 Intervention given at the community level, not at
individual level
 Directed at a given group of patients with specific
conditions
 Randomization done at communities, not at individual
levels

e.g. Impact of health education on EPI programme

 Quasi experimental study
 Qusai = Resembling
Looks like an experimental design but lacks the

key ingredient - random assignment

Typically allows the researcher to control the
assignment to the treatment condition using some
criterion other than random assignment, often by
convenience.

 
 Quasi experimental study (Contd.)

e.g. If we study the effect of maternal alcohol use when the

mother is pregnant, we know that alcohol does harm
embryos.

A strict experimental design would include that mothers were

randomly assigned to drink alcohol. This would be
highly illegal because of the possible harm the study
might do to the embryos.

So what researchers do is to ask people how much alcohol

they used in their pregnancy and then assign them to
groups.
 Quasi experimental study (Contd.)

regarded as unscientific and unreliable by

physical and biological scientists

 very useful method for social scientists 

Can test causal hypotheses
Identifies a comparison group that is as similar as
possible to the treatment group in terms of
baseline (pre-intervention) characteristics
 Meta analysis (Study of studies)
 It is a systematic review of several similarly designed,
small studies on a specific topic
Results of the studies are pooled, summarized and

statistically reanalyzed to get a simple, integrated summary

estimate

 Not exactly a review article rather takes review article one

step further by using revised statistical procedures
 Most commonly methods – Odds ratio (OR) &
Confidence interval (CI)
 Advantages of meta analysis
 Increases statistical power by increasing total sample
size
 Answers questions not originally asked at the beginning
of the study
 Resolves uncertainty when reports of similar study do
not agree
 Observational vs. Experimental Study

1. Based on observation of 1. Based on observation

naturally occurring event of experimentally
created event
2. Nature affects the 2. Researcher intervenes
outcome to affect the outcome
3. Researcher measures 3. Researcher intervenes
and then measures
only, via observation
4. Ethical problem less 4. Ethical problem more
 SHORT OVERVIEW OF STUDY DESIGN

CSS

Historical CS
 Study design
Intervention
No Yes

Observational Experimental

Randomization
Comparison Yes No
Yes group No

RCT CT
Analytical Descriptive
 PROBLEMS
1. Retrospective analysis of 5 yr survival of 200 Ca breast
patients operated in 2008

2. DUB & its effect on fertility: a retrospective analysis over 10 yrs.

3. Prostate volume & post operative outcome following TURP

4. Preoperative hs-CRP & post operative outcome in patients

treated by CABG

5. Plasma BNP & in hospital mortality following AMI

6. Epidemiological evaluation of anthrax

7. Risk factors of LBW

Study Design

1. Historical cohort study

2. Historical cohort study

3. Prognostic cohort study

4. Prognostic cohort study

5. Prognostic cohort study

6. Cross sectional study

7. Case control study or cross sectional study

SAMPLE & SAMPLING METHOD
 Sample and Sampling methods
Population or universe :
entire group of study elements from which data are
collected

Sample :
part of population which represent the population
describing the characteristics of that population
Sampling unit:
every member of sample or unit chosen in selecting
sample
e.g.
 Individuals
 Geographical areas – state, district, village
 Elementary/ Study/Observational unit:

an object or person on which measurement or observation is

made

Sampling units and Study units are sometimes

identical, sometimes different
e.g.

Study of the prevalence of malnutrition among preschoolers

( under 5 years children )
Here,
Sampling unit – may be village

Study unit - under 5 years children of selected

villages
 Parameter
 Summary value of population
 It is always unknown but constant
 Used to represent a certain population characteristics

e.g.

mean stature of all 1st year medical students of

Bangladesh
 Statistic
 Summary value of sample
 Always known but inconstant/varying

e.g.

mean stature of 100 1st year medical students

Sampling :

Process of selection of a number of study units from a

defined study population

Sampling frame (Source list) :

Ordered list of sampling units in the population

e.g.
If 100 students are chosen as sample, then,

Sampling frame -100 students

Sampling unit - each student

 Sampling Techniques

SAMPLING
TECHNIQUES

RANDOM NON-RANDOM
SAMPLING SAMPLING
 Random Sampling
 each sampling unit of total population will have
the equal chance to be included into the sample
 based on random selection
 types :
Simple random sampling
Systematic random sampling
Stratified random sampling
Cluster sampling
Multistage sampling
Multiphase sampling
 Simple Random Sampling
 Simplest form of probability sampling
 Sampling frame is a must
 Sampling units are selected at random by lottery or by
random number table
 Good for small, homogenous and easily available
population
 Systematic Random Sampling

 Sampling frame is needed but not a must

 Sampling units are selected systematically (not randomly)
at constant regular intervals down the sampling frame
 Only 1st sampling unit is selected randomly and then the
rest are selected at a fixed interval
 Good for large, scattered, heterogeneous population
 Stratified Random Sampling
 Total population is divided into homogeneous, non-
overlapping strata by some relevant characteristics (e.g.
sex, age, religion, BMI, income etc.) which can influence
the outcome variables
 Sampling frame of each strata is constructed and
sampling units are selected by simple or systematic
random sampling
 Good for heterogeneous population
 Stratified Random Sampling
Example :

“ Sexual dimorphism in the length of ring and index fingers in 100

medical students”

To make a sample of 20 medical students,

total students are divided into male and female strata

again divided into another strata according to religion

4 strata are generated: Muslim male, Muslim female, Hindu male, Hindu
female

by SRS/SyRS, from each strata, 5 medical students will be selected to

make the sample size 20
 Cluster Sampling

 Selection of groups of study units (clusters) instead of the

selection of study units individually
e.g.
“Nutritional status of medical students of medical colleges of
Bangladesh”
Cluster (Sampling unit) - all medical colleges of Bangladesh
Study unit- medical students
By simple random sampling some govt. medical colleges are
selected and then all medical students of the selected medical
colleges are included in the sample
 Multistage Sampling

 Sampling is done at different stages

 Good when,
o sample frame is not available
o population is too large and widely dispersed
o population is heterogeneous
o study needs a wide area of coverage like community survey
 MULTISTAGE SAMPLING Contd.

Example :

“Effectiveness of EPI programme on under 5 years

children in Bangladesh”

Total population is divided into 1st stage/primary

sampling unit and a sample of such unit is selected by
simple or systematic random sampling
 MULTISTAGE SAMPLING Contd.

Each selected sampling unit is further subdivided into 2nd

stage/secondary sampling unit and then again a sample of
such unit is selected by simple or systematic random
sampling

The procedure continues until the desired stage is reached

 MULTISTAGE SAMPLING Contd.

Country

Division

District

Upazila

Union

Village

Households
 MULTIPHASE SAMPLING

 Part of information is collected from a large sample

and additional information is collected from sub-
samples of whole sample either at the same time or at a
later stage

 Differs from multistage sampling as it is concerned with

similar type of sampling unit at each phase rather than
different types of sampling units used in multistage
sampling
 MULTIPHASE SAMPLING Contd.

Example :

“Prevalence of Dengue fever in a definite area”

In a large sample of population, suspected cases of dengue fever are
identified by signs and symptoms

suspected individuals are allowed to do platelet count

Individuals with low platelet count are detected

IgM test is done in all positive cases to finally identify dengue positive
patients
 NON RANDOM SAMPLING
 Each sampling unit of total population will not have
the equal chance to be included in the sample
 Sampling units are selected by choice or by personal
judgment
 Types :
Convenience sampling
Purposive sampling
Accidental sampling
Quota sampling
Snow ball sampling
 CONVENIENCE SAMPLING

 Sample includes people who are mostly available, easily

accessible and are conveniently selected

 Subjects are selected in a haphazard manner as per

inclusion criteria based on their accessibility and
proximity to the researcher

 Researcher has the freedom to choose whomever he/she

finds within the frame of inclusion criteria
 ACCIDENTAL SAMPLING
 Synonym - Incidental sampling

 People assembled in one place with a common interest

are incidentally surveyed as a sample

 e.g. People attending a seminar or workshop, people in

cinema hall or cricket match etc.
 PURPOSIVE SAMPLING
 Synonym- Judgment sampling
 Researchers judgment is used to select the sample which
he/she thinks to be most typical of the population

◦ e.g. “Correlation of stature and measurements of

foot segments calculated from footprint and foot
outline of adult male Bangladeshi medical
students”

Researcher purposively & according to his/her

judgment selects few students and collects data from
them
 QUOTA SAMPLING

 Total population is divided into relatively

homogeneous , non- overlapping groups (quota)

 Samples are taken from each quota non-randomly by

convenience or purposive sampling

 Similar to stratified sampling but there is no

randomization
 QUOTA SAMPLING Contd.

 e.g.

“People’s opinion about the level of security in a

community”

Total population is divided into different quotas like

politicians, doctors, teachers, students, religious leaders etc.
and then appropriate subjects are selected purposively or
conveniently (not randomly)
For easy understanding:

‘Quota’ can be further divided into more

homogenous ‘Strata’.

e.g.

Quota= Doctor, Engineer etc.

Strata= Male doctor, Feale doctor, Male engineer, Female

engineer etc.
 SNOWBALL SAMPLING
 One eligible person is first identified and then that
person identifies other similar person(s) who is known
to him/her

 Conducted in stages

 Used for hard to find population e.g. drug addicted

persons, HIV and AIDS patients
DATA SUMMARIZATION
&
REDUCTION
 Stages of research
 Planning of research
 Implementation or data collection
 Data management
 Data editing
 Data reduction
 Data presentation
 Data analysis
 Data interpretation &
 Data inference / decision

 Report writing
 Dissemination / publication of the report
 Data editing
Purpose:

 complete the data by correcting the omissions

 to check against illegal entries
 to check for inconsistency of the data
 to check for impossibility of the data
 to discard the meaningless data
 Data editing (Contd.)

Types:
 Validation edits
 Logical edits
 Consistency edit
 Range edit
 Variance edit
 Data summarization/ reduction

 reduction of volume of raw data

 manageable amount without compromising details

 convenient presentation and analysis

 meaningful impression/ summary information of data

Common methods of data summarization

 Tabulation is the common method of data reduction or

summarization

Tabulation can be done in 3 ways:

1. Master table

2. Frequency table

3. Contingency table/ Cross table

 Tables
 displays data in numerical forms in the rows and
column

 display large information in small space

 provides a compact way of presenting large set

detailed information
 Parts of table

 Table number ( in Arabic numerical)

 Title and (subtitle, if any)
 Head note (if necessary)
 Caption or column heading
 Subs or row heading
 Body

 Foot note (if any)

 Source (if not original data)
 Table- (I)
Title :
Head note:

Caption/ Column heading

Subs/ Body
Row
heading
Foot note:

Source:
 Table 4: Breadth of anterior and posterior mitral valve
leaflets in different age groups

Breadth (mm)

Age Anterior Posterior

group Mean ± SD Mean ± SD Probability
value
18-40 years 27.99 ± 3.92 41.23 ± 5.20 0.05*
n=36

41-64 years 33.68 ± 3.56 48.28 ± 3.72 0.001*

n= 30

≥ 65 years 35.33 ± 3.77 50.33 ± 5.27 2.45ns

n=4

P value : * (significant), ns (not significant)

 General principle of table construction

 Simplicity- not more than 3 variables

 Clarity- head note/ foot note

 Self explanatory- without textual any helps

 Directness- only necessary data are included

 Title- complete, clear, concise, to the point

 Source- must be given if not primary data.

 Format- by lines and spaces/ no ditto marks “-”

 Master table
 simple form of tabulation

 shows distribution of observations across several

variables of interest

 each observation is simultaneously cross classified

across variable

 not intended for presentation but only a step

towards deriving various simple or summary table
from it
 Master table (Contd.)

Variable Observations (number)

Male 150

Female 130

Muslim 140

Hindu 80

Christian 40

Buddies 20
 Frequency table
 frequency counts indicate the number of times a data
with particular characteristics occur in a data set

 shows how frequently an event occurs

 traditionally shows frequency counts against

continuous quantitative data

 displays how many scores fall into particular division
of variable
Types of frequency distribution

 Simple frequency distribution

 Grouped frequency distribution
 Relative frequency distribution
 Cumulative frequency distribution
 Cumulative relative frequency distribution
 Frequency table (Exercise)
Examination marks of 30 students:

98,97,48,47,52,58,61,65,70,73,79,84,86,92,93,70,65,8
4,86,86,58,48,52,52,52,58,58,92,92,65,65,73

What to do ?
1st organize the data in ordered array from smallest to
largest as follows:

47, 48, 48, 52, 52, 52, 58,58,58,58, 61,65,65,65,65,

70,70,73,73,79,84,84, 86,86,92,92,92,93,97,98
Frequency table (Ungrouped)

Values Frequency (tally) Values Frequency (tally)

47 | 79 |

48 || 84 ||

52 ||| 86 ||

58 |||| 92 |||

61 | 93 |

65 |||| 97 |

70 || 98 |

73 ||
 Frequency table (Grouped)
ungrouped frequency distribution will be grouped
for summarization
Class interval:
◦ small range of values into which data are
condensed and classified
Class frequency:
◦ number of values in each class
 Class interval (CI)
Range of values into which data are condensed and
classified

Two types:

a) Inclusive type: upper limit included within the

relevant class (I D Good for discrete)
 40-49
 50-59
 60-69
 70-79
Class interval (CI)
b) Exclusive type: upper limit of one class is
excluded for the next class
 40-50
 50-60
 60-70
 70-80
not too large
2 things to remember
about CI not too narrow
E C Good for continuous quantitative
data
Other terms in frequency
Class frequency:
◦ Number of values to each class

Order of CI:
◦ Usually arranged from smallest to largest one

Class limit (CL):

◦ Two ends of each class are regarded as class limit

Class mark:
◦ Mid point of each class
 Cumulative frequency:
Cumulative frequency of an observation is the
sum of all frequency up to that observation.

It is obtained by adding all frequency of previous

observation to that observation.

Useful to know how many values are less than or

more than a certain class.
Relative frequency:
It is the class frequency of a given value expressed
as a percentage of total frequency.

Class frequency
 Relative frequency = X 100
Total frequency

e.g. 20 students (15 MD/MS and 5 M. Phil)

Here relative frequency of MD/MS = 15/20 x100= 75%

and relative frequency of M. Phil = 5/20 x 100= 25%

 References
 Hoque M. , 2014, Dhaka, Bangladesh abc of research
methodology & biostatistics

 Carver R. H. & Nash J. D. , 2012, New Delhi, India Doing Data

Analysis with SPSS

 Anderson D. R., Sweeney D. J. & Williams T. A. , 2011, New

Delhi, India Statistics for Business and Economics

 Bajpai N. , 2010, New Delhi, India Business Statistics