2. When do we do sampling?
Covering
entire population is practically
impossible and the population is infinite.
When the results are required in a short time.
When the area of study is wide.
When resources are limited particularly in
respect of money and trained persons.
When the item or unit is destroyed under
investigation.
3. Why Sampling?
Scientific
approach - Inductive reasoning
Economy - time, money, resources
Quick- procedure is faster
Accurate- results can be accurate
Quality- can be improved
Estimation- adequate and tentative measure
Reliable- error and accuracy
Absence of researcher bias
4. Steps in Sampling
Deciding
universe/population
Is population under study finite or infinite?
Decision about sample Size, Frame
Deciding sampling design (Type &
Procedure)
Calculating sampling error
Statistical generalization…replication
5. What is Universe/Population-Sample?
UNIVERSE: All the individuals/things/events/
documents etc. having designated set of
specifications which a study intends to cover.
POPULATION: All the individuals/things/events/
documents etc. confirming to the designated set
of specifications which the study in particular
covers.
SAMPLE: In relation to population,
representative population, miniature or
aggregate of population.
6. Here is the example….
UNIVERSE:
Children in Mumbai.
POPULATION: Children in the age group of 5
to 10 years, from GMUA, who stay with their
families, and who attend private schools.
SAMPLE: Children residing in the Suburban
areas of Mumbai and attending to Podar,
Jamanabai, Manekji Kooper and Uttpal
Sanghavi Schools.
7. Population versus Sample
Population
= Parameter (N-size, μ-mean, s-SD)
Sample = Statistics (n-size, x- Mean, SD-SD)
Statistics gives estimates about parameter.
A finite subset of statistical individuals defined in
a population is called a sample.
The number of units in a sample is called the
sample size.
The list of the units of sample is sample frame.
10. Types of Sampling
PROBABILITY
[Equal chance,
Estimation of chance]
Simple Random
Systematic Random
Stratified Random
PPS
Area/Cluster
NON-PROBABILITY
[All do not have chance,
No way to
estimate/specify chance]
Accidental/Incidental/
Convenience/Available
Purposive/ Expert
choice/ Judgmental
Quota
Sequential
Snow ball
11. Pre-conditions for
Probability Sampling
Population
is finite
Listing of all the units of the population
Possibility of selection of units at random
Each unit having equal chance of getting
selected
Estimation of chance of selection
Estimation of error in case of non-selection
12. Simple Random
Method:
Chits
Random number tables
Blind folded pointers
Limitations:
Time consuming
Impractical and deviant
Expensive
13. Systematic Random
Method:
Size = Total Number/Required Number
Random beginning at a particular interval
Limitations:
Time consuming
Difficult if high variance in population
At times the cost of data collection is high
14. Stratified Random Sampling
Method:
Formation of strata
Variance among stratum not within stratum
Random subgroups/strata/correlated
categories
Limitations:
Base is the strata, need to know the units
Bigger strata may lead to over representation
15. PPS- Proportionate to Population
Sampling/ Probability Proportional to
Size
Method:
Simple random in stratum
Proportionate to the population in the stratum
Limitations:
Time consuming and expensive
Needs estimates of exact population to decide
proportions
16. Area/Cluster Sampling
Method:
Assumption of homogeneity in the cluster
Usually part of multi-stage design
Limitations:
Deviance or variance within the cluster
Cluster need to be carefully defined
17. Multi-stage Sampling Example
1st: Administrative Ward (Lottery Method)
2nd:Election Ward (Lottery Method)
3rd: Geographic Location for first unit (Purposively)
4th: Identifying Housing society/ Chawl /Flats/Slums
(Random)
5th: Locating household having sample
characteristics (Purposive)
6th: Male and female equal ratio through quota
(Snow Ball)
19. Decision about Sample size
Degree
of accuracy
Extent of variation in population with reference
to key characteristics
Size of the population
Tolerable limits of sampling error
Degree of stratification
20. Calculation of sample size
For a survey design based on a simple random sample,
Formula:
n=
t² x p(1-p)
m²
Where,
n = required sample size
t = confidence level at 95% (standard value of 1.96)
p = estimated prevalence of measure
m = margin of error at 5% (standard value of 0.05)
21. Good sampling design
Adequate
(larger the size better it is)
Accurate & Reliable (least sampling
errors)
Representative (contains all the
properties of the population)
Maximum information about population at
minimum cost, time and human power