I am doing an income analysis of families with children using the supplemental poverty measure. All of my income measures are constant across families. I am using the “spmwt”, but worry that I should restrict my sample so that each family should only be observed once, e.g., where pernum = 1 (though the value of pernum could easily be 2, or 3, as the values of income do not vary by pernum within family).
If I do not restrict the sample, am I not overrepresenting large families?
You are correct that if you do not restrict the sample to a single person per family your estimates will overrepresent large families. However, simply restricting the sample to PERNUM==1 selects one person per HOUSEHOLD rather than one person per FAMILY. Furthermore, the SPM definition of family is different than the official “family unit” definition (see SPMNEWFAM). The SPMWTs were calculated for SPM family units (SPMFAMUNIT) so you to generate representative estimates you would need to select only one person per SPMFAMUNIT ID.
I hope this helps.