NHIS weights


#1

I have merged NHIS data from 2008-2017. Do I need to divide perweight by 10 to get the correct weight?


#2

Yes, you are right that you’ll need to do something to adjust the sample weight to account for the fact that you are pooling data across samples. This is one method that will work relatively well. Technically, you’d want to multiply the sample weight in year i by (sample size in sample i)/(sample size of pooled sample). This will be pretty close to 1/10 in the NHIS data. Additional documentation on this detail can be found on this page under the “Pooling Multiple Years of Data” heading.


#3

Thanks. Am I correct to use perweight as sampling weight? It seems like it corresponds well with wtfa in the NHIS data


#4

If you are performing your analysis at the person-level and using variables for which information is collected for all family members, then yes PERWEIGHT is the correct weight to use.