Proc Mianalyze: How do I pool statistics from Jonckheere-Terpstra Tests?

[Giampaolo]

Hi,

How can I pool the results of the Jonckheere-Terpstra Test from multiply imputed datasets?

Below is my code to generate a datasets with JT statistics from multiple imputations.

Thank you

Giampaolo

proc freq data=mydata;
Table triglycerides*TumorSTAGE/JT noprint;
output out=TT1 JT;
by _imputation_ ;
run;

[SAS_Rob]

Unfortunately there is no direct way to combine non-parametric statistics that depend on the normal distribution for their p-values like the JT test using Proc MIANALYZE.  The theory behind combined estimates requires both a point estimate and a standard error.  

[Giampaolo]

Thank you! What would you recommend then? Report one p value from the imputations? For example the highest to be more conservative?