s-news
[Top] [All Lists]

lme$apVar and PROC MIXED Std. Error of varcomp ests

To: s-news@lists.biostat.wustl.edu
Subject: lme$apVar and PROC MIXED Std. Error of varcomp ests
From: Sven.Knudsen@adeptscience.dk
Date: Thu, 17 Jul 2003 15:52:52 +0200

Dear S group

I am currently evaluating the S-PLUS lme fucntion. The output I try to recreate is from PROC MIXED (SAS version 6.1). As far as I can understand, the two procedures base standard errors of variance components on minus the invers of the Hessian matrix; at lesat that is what the intervals method for lme object does. However, the std. erros differ from the SAS out put when extracted from an lme object via sqrt(diag(lme.obj$apVar)). Am I doing something wrong here?  

Few years ago, I heard a roumer that SAS did compute these values with error - is this what I am discovering here with SAS version 6.1?

Adept Scientific ApS

Sven Jesper Knudsen
Senior Consultant
<Prev in Thread] Current Thread [Next in Thread>