Programming the statistical procedures from SAS

glimmix: computation confidence limits lcl(noblup noilink)

Reply
Occasional Contributor
Posts: 19

glimmix: computation confidence limits lcl(noblup noilink)

title 'final interaction model' ;
proc glimmix data = &lib..&sasdsn plots=meanplot(ilink join cl);
class treatment (ref='mock') genotype (ref='WT') plate ;
model germinated/total = treatment|genotype   /dist=binomial link=logit solution ;
random _residual_ / subject=plate type=cs;
lsmeans treatment*genotype /   cl ; 
output out=out1  pred(noblup noilink)=predicted
				 stderr(noblup noilink)=SE
                 lcl(noblup noilink)=lower
				 ucl(noblup noilink)=upper;
store sasuser.interactionmodel ;
run;
title;

I can reproduce the 95% confidence limits for the lsmeans estimates. This is a t-type confidence interval. However, the lower and upper 95% confidence intervals for the linear predictor(no blups) is narrower. Standard errors are the same as for the lsmeans estimates. Is it for the predicted 5% considered one-sided maybe?

Ask a Question
Discussion stats
  • 0 replies
  • 124 views
  • 0 likes
  • 1 in conversation