Prof. Dr. Md.

Mozammel Hoque
Department of Biochemistry &
Molecular Biology
BSMMU, Dhaka, Bangladesh.
 Research Definition

Scientific and systematic search for knowledge

to
answer certain question
solve a problem

What is problem ?
 Problem

Problem = Expectation – Reality

 Types of Research

1. Quantitative research
2. Qualitative research
3. Basic (pure) research
4. Applied (practical) research
5. Descriptive research
6. Analytical (explanatory) research
 Quantitative Research

• Deals with quantitative phenomenon

• Focus on quantifiable information

e.g. mean blood pressure of doctors of Bangladesh

average height of adult Bangladeshi
 Qualitative Research

⚫ Deals with qualitative phenomenon

⚫ Focus on subjective information
e.g. assessment of attitude, behavior, opinion,
feeling, values, belief of peoples.
 Basic (pure) Research

• Deals with basic processes of a phenomenon.

• Generate new ideas, principles & theories.
• Academically interesting but no immediate utility.

e.g. how malignant cell multiply?

how genes are regulated?
 Applied (practical) Research

• Existing problem oriented research

• Directed to solve an immediate pressing problem
e.g. drug discovery for malignancy
prevention of high infant mortality
 Descriptive Research

• Simply describe a problem from different aspects

• Don’t explore cause-effect relation
e.g. prevalence & distribution of goiter
health care practice of rural peoples
 Analytical (explanatory) Research

• It deals with the determinants of a problem

• Finds cause-effect relationship
e.g. causes & consequences of early MI
causes & consequences of drug addiction
 Common Jargons

• Population : finite / infinite

Entire group of study element.
e.g. all (80,000) doctors of Bangladesh

• Sample
Part (subset) of population. e.g. 4,000 doctors

Contd...
• Parameter: Summary value of population
e.g. mean SBP of all (80,000) doctors,
say 130 mmHg

• Statistic: Summary value of sample

e.g. mean SBP of the sample of 4,000 doctors,
say 125 mmHg

• Inferential statistics: conclusion about the population

parameter based on sample statistic

Contd...
• Variable
Characteristics of population that varies. e.g. BP

• Data
Values or information of a variable measured or recorded.
e.g. 120 mmHg, hypertensive, hypotensive etc.
 Types of Data / Variable

A. Quantitative (numerical): Vary in amount

1. Continuous: In decimal e.g. height, weight
2. Discrete: In integer e.g. family member

B. Qualitative (categorical): Vary in category

1. Nominal: Unranked e.g. sex, religion
2. Ordinal: Ranked e.g. tumor grading
 Types of Variable

A) Independent variable (IV)

exposure, cause, input, risk factor

B) Dependant variable (DV)

outcome, effect, output, response

Obesity (IV) MI (DV)

Contd…
C) Confounding variable

Variable(s) other than IV that can influence the DV.

Obesity (IV) MI (DV)

HTN, DM
(confounder)
 Dichotomous (Binary) Data

Nominal data showing only two categories

e.g. gender, state of pregnancy etc.
 Research Question (RQ)

Definition: queries arising out of the research problem

Example: What are risk factors of early onset MI

Why young peoples are becoming drug addicted?

Is smoking increases the risk of lung cancer?
 Research Hypothesis

▪ Logical & assumed answer to the research question

▪ Examples:
• obesity, DM, HTN, dyslipidemia causes early onset MI
• unemployment, frustration, family breakup cause drug
addiction among young peoples
• smoking increases the risk of lung cancer
 Data Presentation Method

• Table e.g. frequency table, contingency table, cross table

• Figure
a. Graphs: for quantitative data
b. Charts / diagram: for qualitative data

Contd…
Graphs: for quantitative data
Histogram
Frequency polygon
Frequency curve
Line chart
Scattered diagram
Box-plot
Ogive (cumulative frequency curve)
Stem & leaf plot
Contd…
Charts / diagram: for qualitative data
Simple bar diagram
Multiple bar diagram
Component bar diagram
Pie diagram
Pictogram
Map diagram.
ampling
 Sampling Technique

Process of selecting sample from the population

Need representative sample

 Types of Sampling

A. Random (probability) sampling

Done by lottery / random process

• Simple random sampling

• Systematic random sampling
• Stratified random sampling
• Cluster sampling
• Multi stage sampling (area sampling)
• Multi phase sampling

Contd…
B. Non random (non probability) sampling
Done without lottery / random process

• Convenient (accidental) sampling

• Purposive (judgment) sampling
• Quota sampling
• Snowball sampling
 Simple Random Sampling

• For small & homogeneous population

• List all units of population

• Select one by one by lottery
 Systematic Random Sampling

• List of all units of population

• Calculation of sampling interval
• Selection of first unit by lottery
• Selection of next units at every sampling interval

Contd…
Example: N (population) = 60, n (sample) = 20
Sampling frame created
Sampling interval = 3 (60 ÷ 20)

1 10 19 28 37 46 55
2 11 20 29 38 47 56 Select a number from
3 12 21 30 39 48 57 1-3 by SRS. Say it is 2
4 13 22 31 40 49 58
5 14 23 32 41 50 59
6 15 24 33 42 51 60 Start with # 2 & take
7 16 25 34 43 52 every 3rd unit
8 17 26 35 44 53
9 18 27 36 45 54
 Stratified Random Sampling

• Done for heterogeneous population

e.g. population of different religion.

HMCB HMCB H: Hindu

HMCB HMCB M: Muslim
HMCB HMCB C: Christian
HMCB HMCB B: Buddhist

Contd…
• Population divided into homogeneous strata
• From each strata selected by lottery

HHHH MMMM
HHHH MMMM

04 Strata

BBBB CCCC
BBBB CCCC
 Cluster Sampling

• Total population is divided into small clusters

• Clusters are selected by lottery

Garments
Workers in
worker of each garments
Dhaka city factory

Population Cluster
 Multi Stage Sampling

• Sampling done stage by stage

• Stages are determined depending on area

Contd…
Example: Sampling of under 5 yr children of Bangladesh.

Stage Sampling unit (SU)

First Division
Second District
Third Upazilla
Fourth Union
Fifth Ward
Sixth Village
 Multi Phase Sampling

• Sampling done phase by phase

• Phases are created by characteristics of subjects

Contd…
Example:

Prevalence of pulmonary TB in 02 lac population

Clinical features : Phase-1 : 40,000 selected

ESR : Phase-2 : 15,000 selected

X-ray chest : Phase-3 : 10,000 selected

Sputum for AFB : Phase-4 : 8,000 selected

Sputum PCR : Phase-5 : 5,000 selected

 Convenient Sampling

Sample selected considering:

- Easy availability of SU
- Easy accessibility to SU
- Proximity of SU to researcher
e.g. interview of mothers attended in OPD for
treatment of their children.
 Purposive (judgment) Sampling

• Researcher’s judgment is final to select the SU

e.g. tutor selection for your kids

expert opinion on current political scenario
expert opinion on health delivery system
 Quota Sampling

• Done for heterogeneous population

e.g. population of different economic status.

• Population is divided into some homogeneous quota

e.g. rich (quota-1), middle class (quota-2),
poor (quota-3), ultra poor (quota-4).

• From each quota; selected by convenient / purposive

 Snowball Sampling

• Done for hard to find population

e.g. sex worker, drug user, AIDS patients.

• Here one eligible person is first identified & then his / her
help is taken to identify others
Study Design
 What is Study Design ?

Scientific and ethical method of search

To
Collect valid and reliable data.

Contd…
 Example

Say, BP of 130 mmHg

• Data set-1: 98, 99, 100. Not valid but reliable

• Data set-2: 129, 130, 132. Valid & reliable
• Data set-3: 98, 129, 160. Not valid & not reliable

Contd...
• Validity / accuracy
Data close to the fact

• Reliability / precision
Data identical on repeated measurements
 Temporality (Temporal Relation)

• Exposure precede the outcome

• Outcome follow the exposure

Exposure Outcome
(Obesity) (MI)
(Smoking) (Cancer)
 Retrospective Study

• Moves from outcome to exposure

• Deals with pervious data
• Backward looking study
e.g. history of smoking in lung cancer
 Prospective Study

• Moves from exposure to outcome

• Deals with future data
• Forward looking study
e.g. clinical outcome of angioplasty in CAD.
 Longitudinal Study

• Data collected at more than one point of time

• Follow up given on same study subjects
• Mostly prospective but may be retrospective

Onset

Longitudinal & Longitudinal &

Retrospective Prospective
 Cross Sectional Study

• Single time data collection at one point of time

• On a cross section of population
• No follow up & no repeated data collection
e.g. clinical presentation of acute MI: a study on
200 pts done over 3 years
 Observation

Only Observe
Experiment

Do & Observe
What is do? : Intervention (n¯Í‡ÿc)

Something done by researcher to effect outcome

e.g. Drugs, surgery, counselling,

Life style, dietary habit,
Psychotherapy, speech therapy,
Health education,
Music therapy etc.
 Observational & Experimental Study

Cause / exposure Research Outcome

problem
 Types of Study Design

A. Observational study (non interventional)

• Descriptive study
- Case study, Surveillance, Census.
- Cross sectional study

• Analytical study
- Case control study
- Cohort study
- Cross sectional study

Contd…
B. Experimental study (interventional): Analytical

• Clinical trial
• RCT (randomized controlled trial)
• Non randomized controlled trial
 Case Control Study (CCS)
Scenario: Obesity & MI

• Enrolment : by outcome; independent of exposure

➢ Case : subjects with outcome

➢ Control : subjects without outcome

• Compare past exposure in two groups

• Features:
observational, retrospective, analytical

Contd...
 Obese
Case (MI)

Non obese

Obese
Control
(no MI)
Non obese

Time
Onset

Direction of inquiry

Case control study on obesity & MI

 Cohort Study (CS)
Scenario: Obesity & MI

• Enrolment : by exposure; all outcome free

➢ Exposed : subjects with exposure.

➢ Unexposed : subjects without exposure.

• Compare future outcome in two groups

• Features:
observational, prospective / retrospective,
longitudinal, analytical.

Contd...
 Exposed (obese) MI

No MI
Cohort

MI

Unexposed (non obese)

No MI

Time
Onset
Direction of inquire
Prospective cohort study on obesity & MI
Contd...
 Cross Sectional Study (CSS)
Scenario: Obesity & MI

• Enrolment : a cross section of population

(by neither exposure nor outcome)

• Measures exposure & outcome simultaneously

• No follow up; only finds what is happening right now

Contd...
 CSS on obesity & MI

Obese + MI

Non obese + MI Cross section Obese + no MI

of population

Non obese + no MI
 Randomized Controlled Trial (RCT)

Scenario : RCT of a new anti hypertensive drug

• Enrollment : Hypertensive subjects
➢ EXP. gr : Receive intervention
➢ Control gr : Receive std. treatment / placebo

Contd...
Population Exp. gr (new drug) Died

Sampling 50

Cured
Sample 100
Died

50
Control gr (old drug) Cured

Time
Onset
Direction of inquire

Schematic diagram of RCT (old vs new drug)

Contd...
 Randomization

• Random allocation of selected participants in exp. gr. &

control gr. by lottery after sampling
• Protect against bias
 Features of RCT

• Involve human volunteer

• Interventional
• Prospective, longitudinal, analytical
• Randomized
• Controlled
 Blinding (masking)
Ignorance about the treatment status of participants

Blinded
Type
Participants Assessor Researcher
Single blind Yes No No
Double blind Yes Yes No
Triple blind Yes Yes Yes
 Data Summarization (reduction)

Purpose : Impression, presentation & analysis.

Methods : Construction of
1. master table
2. frequency table
3. contingency table
4. cross table
 Grouped Frequency Table
(marks of 500 students)

Frequency is the number of times a data occur in a data set.

Class interval Frequency

0-20 30
20-40 50
40-60 100
60-80 300
80-100 20

* Exclusive type of class interval.

 Grouped Frequency Table
(age of 20 students)

Class interval Frequency

21-22 7
23-24 6
25-26 5
27-28 2

* Inclusive type of class interval.

 One Way or Univariate Table
(contingency table)

Mode of delivery Frequency

Normal

Forceps

LUCS

* Contingency table: frequency table of qualitative data.

 Two Way or Bivariate Table
(contingency table / cross table)

Exposure Outcome (Lung Ca)

(Smoking)
Yes No
Yes
No
 Multi Way or Multivariate Table
(Contingency table / cross table)

Lung cancer No lung cancer

Smoking
Male Female Male Female
Positive
Negative
 Location of Data

Position of a data in a data set compared to other data

 Measures of Location

➢ Measures of central tendency

mean, median, mode.

➢ Percentiles

➢ Deciles

➢ Quartiles

Contd...
Use of the measures of central tendency

➢ Quantitative data : Mean, median & mode

➢ Nominal data : Mode
➢ Ordinal data : Median & mode
 Percentile (P1 to P99) : Centile

➢ 99 imaginary lines that divide data set arranged in

ascending order into 100 equal parts

➢ 5th percentile means, 5% data are below & 95%

data are above that value

Contd…
 P99

P4

P3

P2

P1

Contd…
 Decile (D1 to D9)

➢ 9 imaginary lines that divide data set arranged in

ascending order into 10 equal parts

➢ 5th decile means, 50% data are below & 50% data
are above that value

Contd...
 D9

D4

D3

D2

D1

Contd...
IDR (inter decile range)

D1 to D9

How many data?

 Quartile (Q1 to Q3)

➢ 3 imaginary lines that divide data set arranged in

ascending order into 4 equal parts

➢ 1st quartile means, 25% data are below & 75% data
are above that value

Contd...
 Q3

Q2

Q1

Contd…
IQR (inter quartile range)

Q1 to Q3

How many data?

IDR & IQR………………big: 10-90…………. ???

IDR & IQR………………small: 40-60………. ???

Contd…
Dispersion / spread : distance among the values

* Usually calculated as average deviation of all values

of population from a central value (e.g. mean).

Contd…
 Example

➢ 1st Exam: Q1 = 5 & Q3 = 95 * IQR = 90

➢ 2nd Exam: Q1 = 30 & Q3 = 60 * IQR = 30

* In 2nd exam. Performance is good: Why?

 Measures of Dispersion

Range: Difference between highest & lowest values

Mean deviation (MD) =

 (x - x)
N

Standard deviation (SD) = S =  (x − x) 2

N −1

Contd...
Variance (S2) =
 (x − x) 2

N −1

Inter quartile range (IQR) = Q1 – Q3

Inter decile range (IDR) = D1 – D9

SD
Coefficient of variation (CV) =  100
x
Contd...
Probability
 Probability & P-value

Probability: Chance of occurrence of an event

P-value: Numerical estimate of probability

P-value ranges from 0 to 1

Contd…
P = 0: No chance of the event to occur
e.g. survival after rabies

P = 1: 100% chance of the event to occur

e.g. death of human

P = 0.5: 50% chance of the event to occur

e.g. birth of male baby

P = 0.05: 5% chance of the event to occur

 Probability Distribution (PD)

• Pattern of distribution of a variable in population

e.g. PD of the SBP of doctors of Bangladesh

• Select 1,000 doctors & measure their SBP

Class interval Frequency

100-110 50
110-120 200
120-130 500
130-140 200
140-150 50
Contd…
Histogram of SBP of 1,000 Doctors

Histogram

600

500

400
Frequency

300

200

100

0
100-110 110-120 120-130 130-140 140-150
Blood pressure

Contd…
Probability Distribution of SBP

100 Mean 150

Median
Mode

Contd…
Normal distribution

Mean
Median
Mode

Contd…
 Features of Normal Distribution

▪ Bell shaped & bilaterally symmetrical

▪ Mean, median & mode coincide
▪ About 50% values above & 50% values below the mean
▪ Maximum values lie in the middle around the mean

Contd…
▪ Mean ± 1SD covers 68% values

Mean ± 2SD covers 95% values

Mean ± 3SD covers 99% values

Contd…
Contd…
 Exercise

• mean exam. score (m) = 60 & SD = 5

• 68% students are in the range of 55 to 65

• 95% students are in the range of 50 to 70

• 99% students are in the range of 45 to 75

 Right Skew Distribution
(asymmetric)

➢ Curve skewed to the right.

➢ Higher frequency at lower values.
➢ Lower frequency at higher values

e.g. marks of hard examination.

income of Bangladeshi people.

Contd…
0 100
Positively skewed or skewed to the right
 Left Skew Distribution
(asymmetric)

➢ Curve skewed to the left

➢ Higher frequency at higher values
➢ Lower frequency at lower values
e.g. marks of easy examination
income of American people

Contd…
0 100

Negatively skewed or skewed to the left

 Z-Score

• Deviation of a value from mean in terms of SD.

x - x
• Z-Score =
SD
Example: 500 students, mean score ( x ) = 60, SD = 5.
Z-score of 70 (x) = + 2.
Z-score of 50 (x) = - 2.
Contd...
 Quiz

In an exam; your Z-score = +3

How many students has got marks more than you ?

 Sampling Distribution
Pattern of distribution of sample statistics

Sample 1: mean BP = 120

Sample 2: mean BP = 130
SE
Sample 3: mean BP = 125
GM

SE is the SD among
Sample 90: mean BP = 132 the sample statistics

Contd...
Probability distribution
vs
Sampling distribution

Contd...
Standard deviation (SD)
vs
Standard error (SE)
Sampling Distribution
&
Confidence Interval (CI)
 Title: Mean BP of Doctors of Bangladesh

S1 : n (100), m1 (120), SD1 = 5. PD of sample-1

m = 120, SD = 5
S2 : n (100), m2 (100), SD2.
SD
S3 : n (100), m3 (125), SD3.
m

S90 : n (100), m90 (130), SD90.

Contd...
Sampling distribution of 90 sample means

Here, SEM is the SD of

90 sample means
SEM

GM

Contd...
 GM ± 2 SEM : Include 95% mean
GM ± 3 SEM : Include 99% mean
SEM

GM

95% CI = GM ± 2 SEM
99% CI = GM ± 3 SEM

GM ? ?? SEM ? ? ?

Contd...
• Work on a single large & representative sample
• Find out the mean & SD of that single sample
• Calculate SEM
• Replace grand mean by single sample mean

SD
95% CI = Mean ± 2 SEM
SEM =
99% CI = Mean ± 3 SEM n

Contd...
 Exercise

Mean BP of 100 doctors found 130 mmHg with SD 10

So, SEM = 1

95% CI = m ± 2 SEM = 130 ± 2 = 128-132

99% CI = m ± 3 SEM = 130 ± 3 = 127-133

Hypothesis Testing

What is hypothesis ?

Contd...
Research question (RQ)
Does love makes a man fool?

Research hypothesis (RH)

Love makes a man fool.

Contd…
Research hypothesis is tested indirectly as
NULL HYPOTHESIS

What is null hypothesis ?

Contd…
Research hypothesis (RH)
Love makes a man fool.

Null hypothesis (H0)

Love does not makes a man fool.

Contd…
Research hypothesis (RH)
Lady’s finger reduces blood sugar of gentleman

Null hypothesis (H0) : ??

Contd…
Research hypothesis (RH)
Cucumber enhance facial glow of teenage girl

Null hypothesis (H0) : ??

Contd…
Research hypothesis (RH) & Null hypothesis (H0)
are opposite to each other

if one is true, other one is false

Contd…
 Conceptually, RH is alternative to H0

So, RH is renamed as alternative hypothesis (HA)

RH & HA synonymous

Contd…
How to test research hypothesis (RH)?
By testing of null hypothesis (H0).

To reject H0
OR
To accept H0

Contd…
If H0 rejected : H0 assumed to be false
HA assumed to be true & accepted
Result significant

H0 accepted : H0 assumed to be true

HA assumed to be false
Result not significant

Contd…
How to Test null hypothesis (H0) ??
By
Calculation of the P-value of H0

Contd…
➢ If P-value of H0 = 0…………..H0 is rejected (assumed false)
HA (RH) is accepted; result significant

➢ If P-value of H0 = 1…………..H0 is accepted (assumed true)

HA assumed to be false
Result is not significant

Contd…
P-value = 0 to 0.05 :
H0 rejected, HA accepted; Result significant

P-value = >0.05 to 1 :
H0 accepted, HA rejected; Result not significant

Contd…
P-value calculated by statistical test
P-value determines the fate of H0
Fate of H0 determines the fate of HA
Null Hypothesis (H0) & Judicial Practice

• Hypothesis : accused is guilty

• H0 : accused is not guilty
• Evidence against H0
H0 rejected. Accused punished

* If innocent is punished, H0 wrongly rejected,

false positive conclusion

Contd…
• Hypothesis : accused is guilty
• H0 : accused is not guilty
• No evidence against H0
H0 accepted. Accused acquitted

* If guilty is acquitted, H0 wrongly accepted,

false negative conclusion
 Errors of Hypothesis Testing

Type I (α): Incorrect rejection of H0. False positive.

Type II (β): Incorrect acceptance of H0. False negative.

 Power of Statistical Test

➢ Ability of statistical test to detect a result (effect)

significant when the result is really significant.

➢ Power = 1- β (type II error).

 Risk Estimation: (OR & RR)

Exposure Disease (Ca lung)

(smoking) +ve - ve
Positive a b
Negative c d

Probabilit y of an event to occur

Odds =
Probabilit y of an event not to occur

Odds of outcome in exposed a/b ad

OR = = =
Odds of outcome in unexposed c/d bc
Contd…
 Risk of outcome in exposed a / (a + b)
RR = =
Risk of outcome in unexposed c / (c + d)

Interpretation of OR / RR
• >1 : Risk factor
• 1 : No association
• <1 : Protective factor

Contd…
 Hazard Ratio (HR)

Problem:
04 year follow up study on 60 CRF with DM & 60 CRF
without DM.

Gr. 1: (exposed gr.) - CRF with DM, n=60

Gr. 2: (unexposed gr.) - CRF without DM, n=60

Hypothesis:
CRF patients with diabetes shows increased risk of death.

Contd...
 Interpretation of HR

HR = 1: no association
HR >1 : adverse association
HR <1 : protective association

HR = 2 means, risk of death at any time in exposed group is

two times greater than the risk of death in unexposed group.
 Categories of Test of Hypothesis

1. Parametric test:
Quantitative data with normal distribution
e.g. t-test
Z-test
F-test (ANOVA)
Pearson’s correlation coefficient test etc.

Contd...
2. Non parametric test:
Qualitative data
Quantitative data with skewed distribution
e.g. Chi-square test
Proportion test
Spearman rank correlation test
Fisher exact test
Mann-Whitney U test
 Students ‘ t ’ Test (unpaired)

Preconditions:
• Comparison between two means of two separate groups.
• Quantitative data with normal distribution.
• Unpaired design.
Example:
Comparison of mean BP of between obese & non obese
 Students ‘ t ’ Test (Paired)

Preconditions:
• Comparison between two means of two dependent groups
• Quantitative data with normal distribution
• Paired design
Example:
Effect of dialysis on serum Cr in CRF patients
Effect of exercise on blood pressure
Effect of two drugs given to same patient
 ANOVA (F-test)

Preconditions:
• Comparison among the means of >2 groups.
• Quantitative data.

Example:
Comparison of mean blood pressure among
doctors, engineers, agriculturists & lawyers.
 Proportion Test

Preconditions:
• Compare percentage / proportion between two groups.
• Qualitative data.
Example:
Comparison of the proportion of mortality between new
& old drug
 Chi-square Test (x2-test)

Preconditions:
• Qualitative data of bivariate cross table
• Association between two qualitative variable
Example:
100 patients treated by new drug : 10 patients died
100 patients treated by old drug : 30 patients died
Comment on the new drug
 Performance of a Diagnostic Test

Disease (diagnosed by gold standard test)

Test
+ve - ve
+ve TP (a) FP (b)
-ve FN (c) TN (d)

Test positivity among the

TP(a)
Sensitivit y = disease positive
TP (a) + FN (c)

TN(d) Test negativity among the

Specificit y = disease negative
TN(d) + FP(b)
Contd...
 Disease (diagnosed by gold standard test)
Test
+ve - ve
+ve TP (a) FP (b)
-ve FN (c) TN (d)

TP(a) Disease positivity among the

PPV = test positive.
TP(a) + FP(b)

TN(d) Disease negativity among the

NPV = test negative.
TN(d) + FN(c)
 Correlation

• Relationship between two variables.

• Imply association not causal relation.

 Scatter Diagram

➢ Graphical presentation of the correlation.

➢ Graph is constructed by joining the data points.
➢ Data point : intersection of a definite pair of data.

Contd…
Contd…
Importance of Scatter Diagram

➢ Nature of correlation : positive / negative

➢ Strength of correlation
 Types of Correlation

1. Positive (direct) correlation

Two variables changes in same direction.

2. Negative (indirect) correlation

Two variables changes in opposite direction.

Contd…
Perfect positive correlation.
Both variable changes in same direction in same
proportion.
e.g. temperature & pulse rate.
body weight in pounds & in kg

Contd…
Contd…
Moderate positive correlation.
Both variable changes in same direction but not in
same proportion.

e.g. Height & body weight.

PCV & Hb concentration
Gestational age & fetal weight
BMI & body fat content
Glucose & HbA1c
Income & expenditure

Contd…
Perfect negative correlation (not common).
Both variable changes in opposite direction in same
proportion.

e.g. Pressure of gas & volume of gas

insulin dose & blood glucose concentration

Contd…
Contd…
Moderate negative correlation.
Both variable changes in opposite direction but not
in same proportion.

e.g. Family income & IMR.

TSH & thyroxine
TAG & HDL
Price & demand

Contd…
 No Correlation

e.g. Height & pulse rate

Height & plasma glucose
 Pearson’s Correlation Coefficient Test

• Preconditions:
Correlation between two quantitative variable
e.g. height & body weight

• Calculate correlation coefficient (‘r’ value).

• r- value ranges from -1 to +1.

Contd...
 Interpretation of Correlation Coefficient (r)

• Nature of correlation
➢ If negative : negative correlation.
➢ If positive : positive correlation.
➢ If zero : no correlation / non linear correlation.

Contd...
• Strength of correlation

r – value Correlation (association)

0 No
0.01 to 0.2 Negligible
0.2 to 0.4 Weak
0.4 to 0.7 Moderate
0.7 to 1.0 Strong
1.0 Perfect