Which weight should I use in this particular instance?

Hi all. I’m trying to explore in the 2018 NHIS, but want to go into detail about who is having trouble paying for medical care using demographic questions.

What weight should I use when comparing demographics vs general health questions?

For example, if I wanted to see which race is most likely to worry about medical bills: “race/ethnicity” uses PERWEIGHT, but variables like “Worried about paying medical bills” uses SAMPWEIGHT. I assume I should use PERWEIGHT when breaking down by Demo and SAMPWEIGHT when looking overall.

I ask because the differences in weight can be stark. The other related question is whether to exclude NIU (Not in Universe). My initial thought was to exclude respondents who are NIU since based on the definition it seems like they do not fit the sample. But this again creates stark differences in the data. I’m not sure how to move forward. See this chart for example:

Please advise!! Thanks in advance.

When using multiple variables in a single calculation, which each require different sampling weight variables, a general rule of thumb is to use the sampling weight associated with the most restrictive sub-sample. In your case, the PERWEIGHT variable is the general person-level sampling weight variable and the SAMPWEIGHT variable is the person-level sampling weight associated with the sample-adult questionnaire. The sample-adult questionnaire randomly chooses one adult within each household to answer a longer list of questions. This is also why there are a large number of NIUs for these questions. So, it sounds like that in this case you should use the SAMPWEIGHT sampling weight and exclude NIUs from your calculation.

THANK YOU Jeff!!! This answers our question exactly.