I did a multiple imputation using proc mixed. I got the p-value for linear trend for each of the imputation (I did 20 imputations). But in the combined results I did not get the combined estimate for the test for linear trend.
My independent variable is a continuous variable that I made into 4 categories for a secondary analysis using quartiles as independent variable.
my syntax:
proc mixed data=k.mi;
class catvar_q catvar1 catvar2 catvar3;
model mean_value = catvar_q contvar1 catvar1 catvar2 catvar3 contvar2 contvar3;
lsmeans catvar_q / diff;
by _imputation_;
ods output lsmeans=lsmeans_k (drop=effect);
estimate 'linear' catvar_q -3 -1 1 3;
run; ods listing;
proc sort data=lsmeans_k;
by catvar_q _imputation_;
run;
proc mianalyze data=lsmeans_k;
by catvar_q;
modeleffects estimate; stderr stderr;
run;
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis