Learning SAS? Welcome to the exclusive online community for all SAS learners.

?? Specification of mixed model

Reply
New Contributor
Posts: 2

?? Specification of mixed model

l,eexedPage 401 in the book 'SAS for Mixed Models', second edition (Littell et al 2006), the authors demonstrate a model with PROC MIXED.

The model contains both a random statement and a repeated statement, and both statements allow for heterogeneous variances by specifying effects with the group-option.

The data is available here: http://ftp.sas.com/samples/A59882

I have worked with similar data and a model analogous to Littel's model seem to fit well.

The SAS-code is included below.

Now my problem is: The report, I need to write require model specification in 'statistical notation'.

The fixed effects part is straight forward, but I am uncertain about how to write the random parts.

The random coefficients with the random statement (between subjects) and the heterogeneous residuals with the repeated statement (within subjects).

Somebody help me?

Thanks in advance

Carsten

proc mixed data=scc2 covtest method=ml;

class start forsoegsgrp yktr_dato knr5 koid;

model score_ct =
start*forsoegsgrp forsoegsgrp
knr5 scoregold_ctp50 dekp50 knr5*dekp50 dekp50*dekp50

/ ddfm=betwithin solution outp=pred_sub outpm=pred_avg;

random int dekp50 / subject=koid group=start*forsoegsgrp type=un g gcorr;

repeated / sub=koid group=yktr_dato*forsoegsgrp;

run;

quit;

Frequent Contributor
Posts: 130

Re: ?? Specification of mixed model

You will just need to add an additional subscript (say k) to index over each group, as a unique variance estimate is fitted for each group..
Ask a Question
Discussion stats
  • 1 reply
  • 233 views
  • 0 likes
  • 2 in conversation