I have a question about analyzing complex sample survey data.

In most statistical books on the subject, there are sections that emphasize the need to deal with item missing data and unit missing data. Mostly commonly these are dealt with through weighting and imputation of some sort. I noticed, however that none of my books really talk about how to handle skip patterns in the data. For example, in a health survey, if respondents indicate they are male, they will skip past all the gynecology exam questions. So, my question -- which admittedly may be a dumb one -- is, is it necessary to somehow take into account skip patterns when analyzing this data? If so, how is this done? Is this typically handled using sub-populations (like the SUBPOPN keyword in SUDAAN?) or is it just as simple as coding everyone who skipped the question with a "Not Applicable" value on the variable?

  • 1
    $\begingroup$ hi, assuming the survey weights are reasonable, if you limit the domain to the subpopulation answering the question (the valid domain), then your results should make sense. in the R language, this can be done with y <- subset( x , population_restriction ) $\endgroup$ – Anthony Damico Oct 21 '16 at 12:35
  • $\begingroup$ Hi, @AnthonyDamico, thank you soooo much for your help. I thought that was the case but not sure. This is a perfect answer, so if you want to move it from a comment to an answer (even short answers are "answers" on CV), I'll mark it as correct. Thanks again. $\endgroup$ – StatsStudent Oct 21 '16 at 16:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.