Transcript Slide 1

Module 4 - Session 4.5a
Non - Probability
Sampling Methods
efsa LEARNING PROGRAMME
Common Sampling Designs

Probability sampling
 Random
 Systemic
 Cluster
 Stratified
 Spatial
 Complex
efsa LEARNING PROGRAMME
Common Sampling Designs

Probability sampling
 Random
 Non-probability sampling
 Systemic
 Convenience
 Cluster
 Snowball
 Purposive
 Stratified
 Key informants
 Spatial
 Focus groups
 Quota
efsa LEARNING PROGRAMME
Non-probability Sampling
Overview

Recruit people that are easy to reach or belong
to a group that the assessor is interested

People could be best/worst/typical/unique case

Criteria must be established on how people are
recruited and how characteristics are assessed

Recruitment must be systemically selected

Primarily used for pilot testing
efsa LEARNING PROGRAMME
Non-probability Sampling
Issues

Only report on people that were interviewed

People interviewed may not be representative of
the target population

Cannot make statistically valid generalizations
about target population

Can make statements about the target
population



Range of answers
Inventory outcomes
Presence of a condition
efsa LEARNING PROGRAMME
Convenience Sampling
Overview

Draw sample from people that are available and
easy to reach

Able to get range, inventory, or presence of a
condition

Used for pilot testing such as a random survey

Low cost and quick results

May match and compare convenience samples
to reduce bias
efsa LEARNING PROGRAMME
Snowball Sampling
Overview

Use when sample characteristic is rare

Study subjects recruit other subjects from people
they know

Often used in hidden populations – commercial
sex workers, iv drug users, IDPs, nomads

People who have many friends are more likely to
be recruited

Friends tend to have similar characteristics
efsa LEARNING PROGRAMME
Purposive Sampling
Overview
●
●
●
●
●
Sampling with “a purpose” in mind
Assessor recruits people who fit their criteria
● Best/worse/typical case
● Hard to reach groups – nomads, IDPs
● Live in certain area
Able to quickly reach targeted sample when
sampling for representativeness is not a major
concern
Usually excludes unique or special cases
Often used in rapid assessments
efsa LEARNING PROGRAMME
Key Informants Sampling
Overview
●
People with local knowledge of the target area
and event causing crisis
●
No structured interview
●
Typically reconnaissance in conjunction with
other surveys
●
Identify presence or range of a characteristic
such as a health clinic or distance to a market
●
Information often influences future surveys and
sampling efforts
efsa LEARNING PROGRAMME
Focus Groups Sampling
Overview
●
●
●
●
●
●
●
Combine purposive sampling and self-selection
Usually ask 6-12 people their attitudes,
behaviors, and opinions
Questions are asked in an interactive group
setting where participants are free to talk
Can be used to gain access to cultural and social
groups
Results have high validity
Low cost, fast results
Use – assess wants and needs of clients,
monitor and evaluate services, look at underlying
causes of food insecurity
efsa LEARNING PROGRAMME
Focus Group Sampling
Issues
●
●
●
●
●
●
●
●
●
Assessor has less control over group than in a
one-on-one interview
Small size, inexpensive
Recruitment may be difficult
Need experienced observers or moderators
Members of focus group not large enough to be
representative sample of a population
Assessor may influence results
Assessor are participants of focus group
Focus groups tend to please rather than offer
their own opinions
Focus more on community than individual
attitudes, behavior and opinion
efsa LEARNING PROGRAMME
Quota Sampling
Overview
●
Non-probability equivalent of stratified sampling
●
Assessor identifies stratums and their
proportions
●
Convenience or purposive sampling used to
recruit respondents
●
Select the first people who fit criteria regardless
of what group they represent
●
Key is to select people who match target
population on certain characteristics (age, sex,
etc).
efsa LEARNING PROGRAMME
Quota Sampling
Issues





Selection based upon judgment of interviewer
Only a few characteristics can be simultaneously
considered
No call backs for not-at-home or refusalconversions to complete interviews
Data “looks good” enough to be representative
but there is no randomization and therefore no
idea of the chance of being selected
Data may be “generalizeable” if quota sample
characteristics match target population
characteristics
efsa LEARNING PROGRAMME
a warning…
You may know that results of a survey using non-probability sampling
should not be generalized to a larger population but the reader of
your report may not.
Please consider footnoting this phrase in a report that uses nonprobability sampling:
“This survey used non-probability sampling to select participants.
The results should not be used to make generalizations about a
larger population because the respondents were not randomly
selected and may not be representative of the larger population. In
addition, the chance of being selected for this survey is unknown
and other people in the population being studied had no chance or
an unknown chance of being selected for this survey.”
efsa LEARNING PROGRAMME