DR. MD.

HABIBUR RAHMAN
MBBS, FCPS P1 (MEDICINE)
BCS (RECOMMENDED)
• Population: Entire group of study elements. Eg. All
doctors in bd
• Sample: part (Subset) of population e.g. 4000 doctors

• Parameter: Summary value of population. E.g. E.g.

mean SBP All doctors in bd
• Statistics: Summary value of sample. E.g. mean SBP of
4000 doctors is 125 mm Hg
• Variable: Characteristic of population that varies. E.g.
BP

• Data: Value or information of a variable measured or

recorded. E.g. 120 mmHg
 TYPES OF DATA
Quantitative (Numerical): Vary in amount
Continuous: in decimal e.g. height (5.2 feet), weight (41.2 kg)
Discrete: In integer e.g. family member (1/2/3)

Qualitative (Categorical): Vary in category

Nominal: Unranked e.g. sex, religion, BMDC reg. no.
Ordinal: Ranked e.g. Tumor grading, GCS score, APGAR score
 TYPES OF VARIABLE
Independent variable (IV): exposure, cause, risk factor, input. E.g.
smoking
Dependent variable (DV): Outcome, effect, output, response. e.g.
Lung cancer
Confounding variable: Variables other than IV that can influence DV.
E.g. HTN causes MI and DM also causes MI in a patient.
Intervening: Independent variable influences dependent variable
through this variable. E.g. Dyslipidemia causes MI through
atherosclerosis.
 SCALE OF MEASUREMENT
Normal scale
Ordinal scale qualitative

Interval scale - thermometer

quantitative
Ratio scale – Body weight, height measurement
 DATA PRESENTATION
QUALITATIVE DATA: Diagram
QUANTITATIVE DATA: Graph
 GRAPHS
For quantitative data
1. Histogram
2. Frequency polygon
3. Frequency curve (Normal distribution curve)
4. Scatter/ dot diagram
5. Line chart/ graph
 CHARTS/ DIAGRAM
Qualitative
1. Simple bar diagram
2. Multiple bar diagram
3. Component bar diagram
4. Pie diagram
5. Pictogram
6. Map diagram
 SAMPLING
Random/ probability:
1. Simple random
2. Systematic random
3. Stratified random
4. Cluster random
5. Multistage
6. Multiphase
 Non-random/ non-probability
1. Convenient (accidental) sampling
2. Purposive
3. Quota sampling
4. Snow ball
 SIMPLE RANDOM SAMPLING
For small and homogenous population e.g. Lottery

STRATIFIED RANDOM SAMPLING: Heterogeneous population e.g Hindu,

Muslim, Christian….

MULTSTAGE SAMPLING (WHOLE COUNTRY)

Sampling done stage by stage
Division > District > Upazilla > Union > Village

QUOTA SAMPLING: Done for heterogenous sampling

STUDY DESIGN
 RETROSPECTIVE STUDY
Outcome to exposure
Deals with previous data
E.g. History of smoking in lung cancer
 PROSPECTIVE STUDY
Exposure to outcome
Deals with future data
Forward looking study
E.g. smoking causes lung cancer
 LONGITUDINAL STUDY
Data collected at more than one point of time
Follow up given on same study subjects
 CROSS SECTIONAL STUDY
Single time data collection at one point time
No follow up and no repeated data collection
 TYPES OF STUDY DESIGN
Observational study (not interventional)
Descriptive:
Case study,
case series,
surveillance,
census,
cross sectional

Analytical: Case control, Cohort study, Cross sectional

 Experimental study (Interventional): Analytical
Clinical trial
RCT
Non RCT
 DIFFERENCES

Observational Experimental
No intervention Intervention given
ess ethical constrain More ethical
More bias Less bias
 CASE CONTROL STUDY
Case- Subject with outcome Control- Subject without outcome
Advantages:
Good for rare diseases
Good for diseases of long latent period
Little attrition period , Less ethical constrain
Easy to do and less time needed
Disadvantages:
Bad for rare exposure
Can not measure incidence/ prevalence
Does not prove causality, prone to bias
 COHORT STUDY
Scenario: Obesity & MI ; Hypothesis: Obesity causes MI
Advantages:
Good for rare exposure, Less prone to bias
Can measure incidence
Can prove causality

Disadvantages:
Costly and long time needed
Attrition (Loss of follow up)
Not good for rare diseases
 CROSS SECTIONAL STUDY
Scenario: Obesity & MI ; Hypothesis: Obesity causes MI
Advantages:
Quick and easy, less cost
Can measure prevalence
Generate hypothesis

Disadvantages:
Can not prove causality
Not goo for rare disease/ rare exposure
Not good for fatal disease/ disease of short duration
 RCT
Inventioned study
Scenario: RCT with a new anti-hypertensive drug
Hypothesis: New drug is better than old drug
Advantages:
Reliable and valid, Can prove causality
Good control of confounders
Disadvantages:
Expensive and time consuming, ethical constrain
Good internal validity, poor external validity
Non-compliance and attrition
 MEASURES OF LOCATION
Measures of central tendency: Mean, Median, Mode
Percentiles
Deciles
Quartiles
Mean: Arithmetic mean
Median: Middle most value when data are arranged in ascending or
descending order
Mode: Most occurring value

Percentile: Value is divided into 100 equal parts. E.g. 5 percentile

means 5% data are below this value and 95% data is above this value
Decile: Value is divided into 10 equal parts. E.g. 4th decile means 40%
data is below this value and 60 % data above.
Quartile: Value are divided into 4 equal parts.
 MEASURES OF DISPERSION/ SPREAD
Dispersion/ spread:
1. Range
2. Mean deviation
3. Standard deviation (SD): This is the square root of variance (mostly
used )
4. Variance – most
5. Inter quartile range
6. IDR
7. Coefficient of variation
 DATA DISTRIBUTION
TYPES:
1. Normal or Gaussian distribution
2. Skewed distributiob
3. Log-Normal distribution
4. Binominal distribution
5. Poisson distribution
CHARACTERISTICS OF NORMAL DISTRIBUTION
CURVE
Bell shaped
Symmetrical
Mean = median = mode
Total area of curve 1
50% data above and 50% data below the value
Mean ±1 SD covers 68% values
Mean ±2 SD covers 95% values
Mean ±3 SD covers 99.7% values
 SKEWED DISTRIBUTION
Asymmetrical
Positively skewed: Mean>median>mode. Mean & median are to the
right of the mode
Negatively skewed: Mean<median<mode. Mean & median are to the
left of the mode
Mean is towards longer tail, mode towards shorter tail
 PROBABILITY
Chance of occurrence of any event by chance.
Value of P is in between 0 & 1
P= 0 means zero chance of occurrence e.g. survival after rabies.
P= 1 means 100% chance of occurrence e.g. Death of animal
Cut off point : 0.05
Significant result: if P value <0.05
 HYPOTHESIS
NULL: There is no difference between control and experiment value
ALTERNATIVE: There is a difference
P value <0.05 means chance of error is less and Null hypothesis can
be rejected and alternative hypothesis is accepted. Result is significan
 SAMPLING ERROR
Type 1 error/ alpha: equivalent to false positive. Incorrect rejection o
null hypothesis

Type 2 error: equivalent to false negative. Incorrect acceptance of

null hypothesis.
 TEST OF HYPOTHESIS/ SIGNIFICANCE
Parametric: For quantitative data (normally distributed data)
t-Test (Student’s t-Test)
F-test (fisher’s test or analysis of variance/ANOVA)
Z-test (Proportion test)
Regression test
Non-parametric: for qualitative data (asymmetrically distributed data
Chi-squared test
Fisher exact probability test
Mann-Whitney test
Z-test
 t-Test
Find out the significance between two mean.
Criteria:
Parametric
Quantitative data
Normally distributed data
Random sampling
Sample size < 30
 Z-TEST
Find out the significance between two proportion
Criteria:
Parametric or non-parametric
Quantitative data
Normally distributed data
Random sampling
Sample size > 30
 CHI-SQUARED TEST
Find out the significance between two mean.
Criteria:
Non-Parametric
Qualitative data
Asymmetrically distributed data
No mean, median or mode
Random sampling
Sample size < 30
 F-TEST OR ANOVA
Find out significant difference between 2 or more data.
Sensitivity: Detects true positive. E.g. ANA in SLE
Specificity: Detects true negative e.g. Anti-dsDNA in SLE
True positive:
True negative:
Positive predictive value
Negative predictive value