Sampling Techniques and Survey Methodology
Core concepts
Know probability sampling methods: simple random sampling, systematic sampling, stratified sampling, cluster sampling, and multistage sampling, plus their estimators and variances. Understand sampling and non-sampling error, sampling frames, coverage, nonresponse, and weighting. Be familiar with the design effect, intra-class correlation in cluster samples, optimal allocation in stratified sampling (Neyman allocation), and ratio and regression estimators. Understand questionnaire design pitfalls and the difference between cross-sectional and longitudinal surveys.
Traps
- Confusing stratified sampling (sample within every stratum) with cluster sampling (sample whole clusters).
- Ignoring the design effect when computing standard errors for complex designs.
- Treating nonresponse as random when it is systematically related to the variable of interest.
- Using proportional allocation when variances differ greatly across strata; Neyman allocation is more efficient.
- Confusing sampling error (reducible by increasing n) with non-sampling error (not necessarily reduced).
- Assuming a systematic sample is always equivalent to a simple random sample when the list has periodicity.
Worked example
Two strata with sizes N1=600, N2=400 and standard deviations s1=10, s2=20. Neyman allocation gives n1/n2 = (N1 s1)/(N2 s2) = 6000/8000 = 0.75, so with n=70, n1=30 and n2=40. Rule of thumb: allocate more sample to strata that are larger and more variable; proportional allocation ignores variability and is less efficient when strata differ in spread.