Hello, I am some what new to SAS and very new to using Proc Glimmix. I recently took a multilevel modeling course and it is new to me. I am working with sample weights and I am confused. I am having a hard time understanding my output. I am also wondering if my code is correct. I wanted to run a two level model. I have state level variables (level two) and I wanted to understand their relationship to Postpartum Depression (level one). Postpartum Depression is binary (yes or no). I also have other health indicators at level one and state level poverty at level 2. Maybe I am confused..but I don't think my code is correct. I would like to have the odds ratios in the output as well. My IV (state level variables at level 2) are categorized as 0=low, 1=med, and 2=high. I am really stuck and would appreciate any help! I wanted to do at random effects and fixed effects ( for the state level variables). . I am wondering if it correct and also if I wanted to do all fixed effects what should I enter. I an introductory course and I am still learning. However, the course was not in SAS, so I am really confused. Here is my code proc sort data= out1; by FIPS; run; proc summary data=out1 print; by FIPS; var wtanal; output out=intermediate uss=sumsqw sum=sumw n=nj; run; proc means; run; data LaurenFinal; merge out1 intermediate; by FIPS; aw=wtanal/(sumw/nj); bw=wtanal/(sumsqw/sumw); run; proc means; run; proc contents data=out1; run; proc freq data=LaurenFinal; tables state FIPS; run; /*method A*/ Proc glimmix noclprint data=LaurenFinal method=quad; class FIPS; model PostpartumDepression= /dist=binomial link=probit obsweight=aw solution; random intercept /sub=FIPS; covtest 'var (FIPS)=0' 0.; run; /*Method b*/ Proc glimmix noclprint data=LaurenFinal method=quad; class FIPS; model PostpartumDepression= /dist=binomial link=probit obsweight=bw solution; random intercept /sub=FIPS; covtest 'var (FIPS)=0' 0.; run; THIS IS WHAT I HAVE NOW, BUT I DON'T THINK IT'S RIGHT: Title' PPD overall BW Incarceration model method A2 Random'; Proc glimmix noclprint data=LaurenFinalPhD method=quad; class FIPS; class BWIncRatioAVG_3/REF=First; model PostpartumDepression (event='1')= /dist=binary Oddsratio link=logit obsweight=aw solution; random intercept BWIncRatioAVG_3 /sub=FIPS CL; covtest 'var (FIPS)=0' 0.; run; Title' PPD overall BW Incarceration model method b Random'; Proc glimmix noclprint data=LaurenFinalPhD method=quad; class FIPS; class BWIncRatioAVG_3/REF=First; model PostpartumDepression= /dist=binomial Oddsratio link=logit obsweight=bw solution; random intercept BWIncRatioAVG_3 /sub=FIPS CL; covtest 'var (FIPS)=0' 0.; run; /*Use fixed for models. this is the crude model*/ Title' PPD overall BW Incarceration model method A2 fixed'; Proc glimmix noclprint data=LaurenFinalPhD method=quad; class FIPS; class BWIncRatioAVG_3/REF=First; model PostpartumDepression (event='1')=BWIncRatioAVG_3 /dist=binary Oddsratio link=logit obsweight=aw solution; random intercept /sub=FIPS CL; covtest 'var (FIPS)=0' 0.; run; Title' PPD overall BW Incarceration model method b fixed'; Proc glimmix noclprint data=LaurenFinalPhD method=quad; class FIPS; class BWIncRatioAVG_3/REF=First; model PostpartumDepression (event='1')=BWIncRatioAVG_3 /dist=binary Oddsratio link=logit obsweight=bw solution; random intercept /sub=FIPS CL; covtest 'var (FIPS)=0' 0.; run; Thank you so much!!
... View more