Repeated and Random statement in Proc Mixed

fatemeh — Fri, 24 Nov 2017 17:17:18 GMT

hello all,

I have a question about proc mixed when having both fixed and random effect .when i put both random and repeated statement i get warning because of infinite likelihood when i want output the FitStatistics. repeat statement is for R side and random is for G side but when i put both repeat and random we suppose within subject errors have no correlation the R matrix should be diagonal (σ^2) I. so to prevent this error should i remove type=cs statement from proc mixed ? bellow is my code. thanks for any help.

proc mixed data=data;
class Subject time treat ;
model response=time|treat / ddfm=kr;
random Subject(treat);
repeated time / type=CS subject=Subject(treat) ;
ods output FitStatistics=Fitcs(rename=(value=cs))
Dimensions=Parmcs(rename=(value=Numcs));

run;

Re: Repeated and Random statement in Proc Mixed

sld — Sat, 25 Nov 2017 19:12:20 GMT

Both

random subject(treat);

and

repeated time / type=cs subject=subject(treat);

fit the same covariance structure and to use both in the same model is redundant, causing the model to be overspecified. Use just one or the other.

See the paper by Littell et al. (2000)

http://onlinelibrary.wiley.com/doi/10.1002/1097-0258(20000715)19:13%3C1793::AID-SIM482%3E3.0.CO;2-Q/abstract

for a good description of the more common covariance types and why you would or would not include both random and repeated statements depending on type.

topic Re: Repeated and Random statement in Proc Mixed in Statistical Procedures

Repeated and Random statement in Proc Mixed

Re: Repeated and Random statement in Proc Mixed