PURPOSIVE/THEORETICAL SAMPLING
Representative and Random Sampling
Purpose = generalization from sample to population
Better inferences with homogeneous population strata
(e.g., North Shore residents with incomes b/w $75,000-$85,000)
Stratification = strata that are more alike contextually
Purposive Sampling
Begins with the assumption that context is critically important
Doesn't assume there's a point at which context can be ignored
Each context must be dealt with in its own terms
Purposive Sampling Techniques (Purposes of Samplin:
Extreme cases -- particularly unusual, troublesome, or enlightening
Typical cases -- avoid rejecting info as arising from deviant cases
Critical cases -- permit maximum application of information (iffthen)
Political cases -- _attract/deflect attention to/from the inquiry
Convenience -- save time, effort, resources
Maximum Variation -- unique variations in adapting to conditionsGoal of Sampling
Maximum information on site, members, and nature and effects of constructions,
relations, processes, and products
Not after similarities that can be formed into generalizations
Want detailed and specific understanding of unique character of context
Maximum information as basis for emergent design and grounded theory
Verify information collected and analyzed
Emergent Sampling Design
Use different techniques at different times (as purposes change)
Typical cases alert you to need for extreme cases which uncover sensitive issue
for which you need political cases
Use multiple techniques at same time
Typical ESA in an extremely rural area in a crit
Selection of Sampling Units
Primary Site selected using six techniques, alone or combined, and conceptualized
in terms of levels of organization, participation, and/or (political) interest
Secondary Sites selected using six techniques, alone or combined, at each level
Serial Selection in multiple-site studies: first site studied in relative depth before
selecting other sites using six techniques, alone or combined
Serial Selection of participants: select and interview/observe some participants
before purposively selecting others, based on insights from initial or earlier
interviews/observationsSnowball Sampling
Continuous participant nominations of similar and different participants, settings,
and documents/records
Continuous Focusing of Sample
At first, everything is relevant (selection bias = different from what you have)
As problem emerges, sampling becomes more problem-relevant
(select bias = same as per more depth on problem-relevant matters)
Redundancy
End point, given resource constraints
Planning Considerations
Initial elements selection:
Logical — all stake holding groups across levels of organization, participation,
and interest
Nominations by elites, gatekeepers, informants
Orderly emergence of sample via serial selection, snowballing, problem focusing