Hello!
I have a query regarding the NIU (Not in Universe) cases for certain variables in IPUMS NHIS 2015. For my project, I’ve restricted the dataset to individuals aged 18 and older.
The variables in question are: SAWGEN, SAWSPEC, DELAYAPPT, DELAYWAIT, SMOKESTATUS2, MOD10FWK, VIG10FWK, and STRONGFWK. Each of these variables shows 44,437 NIU cases, which is a little puzzling because according to the documentation, the universe for SAWGEN, SAWSPEC, DELAYAPPT, and DELAYWAIT includes sample adults 18+ and sample children under 18, while the universe for the remaining four variables includes only sample adults 18+.
Could you please explain why these 44,437 NIU cases are appearing in my data? Or, could you please direct me to resources that might help clarify this issue? Thank you! 
I tabulated the variables you listed using the online data analysis tool for the 2015 NHIS data. I looked at the unweighted counts of individuals who were NIU for each variable. I got the following results:
- SAWGEN, SAWSPEC, DELAYWAIT, and DELAYAPPT: The universe is sample adults age 18+ and sample children under age 18 and there are 57,826 NIU individuals
- SMOKESTATUS2, MOD10FWK, VIG10FWK, and STRONGFWK: The universe is sample adults age 18+ and there are 70,117 NIU individuals
You stated you restricted your analytical sample to individuals age 18+. By doing that, you will not include in the count any NIU individuals who are under age 18. The presence of individuals under age 18 in the NIU group is the difference between the universes of the two sets of variables.
You may want to review the variance estimation page for NHIS, particularly the sections on subsetting your data, for information on how to obtain accurate standard errors with a subset of the NHIS data.
1 Like
Thanks again, Isabel. I will re-run the analysis to check.