topic Re: PROC LOGISTIC - DIFFERENT WEIGHTS FOR DIFFERENT VARIABLES in Statistical Procedures

PROC LOGISTIC - DIFFERENT WEIGHTS FOR DIFFERENT VARIABLES

jazambuja — Sat, 07 Mar 2020 21:06:28 GMT

Hi team,

I am currently working with the HSLS:09 survey data. This dataset combines data from four mains surveys (students, parents, math teachers and science teachers). Each one of these surveys has a different analytical weight associated with them. My logistic analysis uses variables from all these surveys. Do you know how can I incorporate these different survey weights in the PROC SURVEYLOGISTIC or another procedure?

Can anyone help me with this? I am really struggling.

Re: PROC LOGISTIC - DIFFERENT WEIGHTS FOR DIFFERENT VARIABLES

tellmeaboutityo — Mon, 09 Mar 2020 20:35:45 GMT

I've never personally done it, but I believe you can use proc glimmix with logistic assumptions.

proc glimmix data= survey; 
	class dv; 
	model dv = iv1 iv2 / dist=binomial link=logistic obsweight=weightvariable oddsratio solution; 
run;

If binomial doesn't work, it might be bernoulli, which will give the same result. I can't remember if glimmix allows both.