Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks

Sampling People, Records,

& Networks
Jim Lepkowski, PhD
Professor & Research Professor Emeritus
Institute for Social Research, University of Michigan
Research Professor,
Joint Program in Survey Methodology, University of Maryland

Copyright James M. Lepkowski 2016

Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
Unit 1
• 1 Research • Unit 1: Sampling as a research tool
designs ... • Lecture 1 – Research design & sampling
• 2 Surveys • Lecture 2 – Surveys & sampling
• 3 Why sample? • Lecture 3 -- Why sample at all?
• Lecture 4 – Why might we randomize, and how do we do it?
• 4 Why
randomize? • Lecture 5 – What happens when we randomize?
• Lecture 6 – How do we evaluate how good the sample is?
• 5 Types of • Lecture 7 – What kinds of things can we sample?
sampling
• 6 Evaluating • Unit 2: Mere randomization
samples
• Unit 3: Saving money
• 7 Units
sampled • Unit 4: Being more efficient
• Unit 5: Simplifying sampling
• Unit 6: Some extensions & applications

Copyright James M. Lepkowski 2016

2
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
Unit 1
• 1 Research • Unit 1: Sampling as a research tool
designs ... • Lecture 1 – Research design & sampling
• 2 Surveys • Lecture 2 – Surveys & sampling
• 3 Why sample? • Lecture 3 -- Why sample at all?
• Lecture 4 – Why might we randomize, and how do we do it?
• 4 Why
randomize? • Lecture 5 – What happens when we randomize?
• Lecture 6 – How do we evaluate how good the sample is?
• 5 Types of • Lecture 7 – What kinds of things can we sample?
sampling
• 6 Evaluating • Unit 2: Mere randomization
samples
• Unit 3: Saving money
• 7 Units
sampled • Unit 4: Being more efficient
• Unit 5: Simplifying sampling
• Unit 6: Some extensions & applications

Copyright James M. Lepkowski 2016

3
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample
• Accuracy • Unit 1: Sampling as a research tool
• Lecture 1 – Research design & sampling
• Probabilities
• Lecture 2 – Surveys & sampling
• Frames • Lecture 3 -- Why sample at all?
• Techniques • Lecture 4 – Why might we randomize, and how do we do it?
• Deficiencies • Lecture 5 – What happens when we randomize?
• Lecture 6 – How do we evaluate how good the sample is?
• Complex
design
• Lecture 7 – What kinds of things can we sample?
• Unit 2: Mere randomization
• Unit 3: Saving money
• Unit 4: Being more efficient
• Unit 5: Simplifying sampling
• Unit 6: Some extensions & applications

Copyright James M. Lepkowski 2016

4
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • During conceptualization, a researcher considers the RELEVANT
• Accuracy POPULATION for evaluating the theory/hypothesis
• Probabilities
• Frames
• Techniques • In designing the data collection, the researcher has two concerns
• Deficiencies
in mind:
• External validity
• Complex
design • Cost/benefit calculations for the overall cost of the study
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample A census involves an enumeration of a population. When the
• Accuracy population is large:
• Probabilities
• Frames
• Techniques 1. It is costly
• Deficiencies
2. It is time consuming
• Complex
design 3. It may not be feasible with complete precision
(US Census as an example)
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample A sample involves a selection of a representative subset of a
• Accuracy population in order to draw inferences to the population
• Probabilities
• Frames
• Techniques Collecting data from a sample of a large population is FAR
• Deficiencies LESS costly and FAR LESS time consuming
• Complex
design
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample Recruitment directly – volunteer samples
• Accuracy
• Probabilities
• Frames Lists, selection, & then recruitment
• Techniques
• Deficiencies
Lists, selection, recruitment, & nonresponse
• Complex
design
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Because of the cost savings, sampling allows a researcher to
• Accuracy devote
• Probabilities • More resources to the collection of more data (variables)
• Frames • The reduction of error in measurement (reliability and validity)
• Techniques • Better coverage of the units of analysis
• Deficiencies
• Complex
design • This fits in with what is called a Total Survey Error perspective
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Non-probability sampling
• Accuracy • Haphazard, convenience, or accidental sampling
• Probabilities • Purposive sampling or expert choice
• Frames • Quota sampling
• Techniques • Substitution (for nonresponse)
• Online panels
• Deficiencies
• River sampling
• Complex
design

10
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Probability sampling
• Accuracy
• Simple random selection
• Probabilities
• Frames
• Stratified selection
• Techniques • Cluster samples
• Deficiencies • Systematic samples
• Complex • More complex samples: probabilities proportionate to size
design

11
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • List frame
• Accuracy
• Area frame
• Probabilities
• Frames • Problems
• Techniques • Missing elements
• Duplicate listings
• Deficiencies
• Clusters
• Complex
design • Blanks or ineligibles

12
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Simple random sampling
• Accuracy
• Systematic sampling
• Probabilities
• Frames • Stratified sampling
• Techniques • Proportionate allocation
• Deficiencies
• Disproportionate allocation

• Complex
design

13
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Cluster sampling
• Accuracy
• Two-stage sampling
• Probabilities
• Frames • Probability proportionate to size sampling
• Techniques • Stratified probability proportionate to size sampling
• Deficiencies
• Multistage sampling
• Complex
design
• Multiple phase sampling

14
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Nonresponse
• Accuracy • Total/unit
• Probabilities • Item
• Frames • Noncoverage
• Techniques • Compensation: weighting
• Deficiencies • Unequal probabilities
• Complex • Nonresponse
design • Noncoverage (poststratification)
• Make the sample distribution conform to known population distribution

15
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Complex designs typically involve one or more of …
• Accuracy • Stratification
• Probabilities • Clusters
• Frames • Weights
• Techniques • Estimation becomes complex
• Deficiencies • Even a simple mean or proportion requires non-standard techniques
• Complex
design

16
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
• Census or
sample • Standard software cannot handle complex sample designs
• Accuracy correctly
• Probabilities
• Estimating precision becomes more complex as well
• Frames
• Techniques • Methods of variance estimation must be considered
• Taylor series approximation
• Deficiencies
• Balanced or Jackknife repeated replication
• Complex
design • Computer software available for these methods
• Requires stratum, cluster, and weight on each sample record

17
Survey Data Collection & Analytic Specialization Sampling People, Records, & Networks
Unit 1
• 1 Research • Unit 1: Sampling as a research tool
designs ... • Lecture 1 – Research design & sampling
• 2 Surveys • Lecture 2 – Surveys & sampling
• 3 Why sample? • Lecture 3 -- Why sample at all?
• Lecture 4 – Why might we randomize, and how do we do it?
• 4 Why
randomize? • Lecture 5 – What happens when we randomize?
• Lecture 6 – How do we evaluate how good the sample is?
• 5 Types of • Lecture 7 – What kinds of things can we sample?
sampling
• 6 Evaluating • Unit 2: Mere randomization
samples
• Unit 3: Saving money
• 7 Units
sampled • Unit 4: Being more efficient
• Unit 5: Simplifying sampling
• Unit 6: Some extensions & applications

Copyright James M. Lepkowski 2016

18