I am using several datasets that can be svyset. In each of these, I am interested in only observations that identify as religious. I understand that dropping observations is inappropriate for svy: data. I also understand the command for subpop() when examining survey data. My basic problem is that it is going to be tedious to go and re-specify subpop for each of the analyses that I'm doing.
I was thinking of "cheating" and using complete case analyses so that I recode a minor covariate as 'missing' for the group who I do not want to consider. This way they are excluded via complete case analyses and I have to change six lines of syntax instead of ~100.
Instead of:
Code:
recode relig (4=0) (1/3=1) (5/13=1), gen(filter) svy(filter): regress y x minorcovariate
Code:
replace minorcovariate=. if filter==0 svy: regress y x minorcovariate
I'm guessing that what I've done is problematic because it seems easy. Moreover, if it is problematic, why is any complete-case analysis allowed with svyset? Observations are always excluded. I'm betting there is a reason, but I would love to have some conceptual insight into it.
Cheers,
David.
0 Response to Conceptual/statistical question about supop().
Post a Comment