In IHIS, I am looking at non-cigarette tobacco use for 2000 and 2005, i.e. variables BIDIEV, PIPEV, CIGAREV etc.
As well as the “NIU”, “don’t know” and “refused codes”, there are a considerable (and consistent) number of occurrences of code 8=“not ascertained”. I assume this is due to only a subsample answering this batch of questions.
I am planning to use crosstablualtion to generate some fairly simple tables showing prevalence of usage by various other factors. Would it be appropriate to omit those “not ascertained” codes from the analysis along with the NIUs?
Is the SAMPWEIGHT still the correct weighting to use?