http://www.nabble.com/file/p24211204/repeated.csv repeated.csv
Dear all,
I'm currently trying to replicate some Proc Mixed results using lme() and
have a curious result I can't explain.
The dataset is a repeated measures example where patients (each on one of
several treatments) are measured over a number of days. Putting aside issues
of the error covariance structure I'm using the simple compound symmetry.
The SAS code I'm using is:
Proc mixed ;
class dose day animal ;
model blood=dose|day / htype=1,3;
repeated day / subject = animal type = cs;
run;
and the lme code is
testmain<-lme(blood~dose*day, random=~1|animal, data=sasdata, na.action
(na.omit))
anova(testmain)
The overall test of the fixed effects agree as expected, but when I tried
Blood2 (which has missing data) only the Day*Dose interaction agrees. I
tried sequential and marginal options in the anove.lme code but to no
effect. I suspect this is something to do with the "non-orthogonality"
induced by the missing data but I am not sure.
Not being an expert in this area I was wondering if anyone knew why I'm
seeing these differences and if I can tweak the R code to get agreement?
Any thoughts would be most appreciated!
Simon
--
View this message in context:
http://www.nabble.com/lme-gives-different-results-to-SAS-Proc-Mixed-tp24211204p24211204.html
Sent from the R help mailing list archive at Nabble.com.
