Hello,
I am using SAS 9.4 TS Level 1M3. I am trying to bootstrap confidence intervals around variance partitioning coefficients for hierarchical data. My model for calculating my VPC is:
Proc mixed data= lowserum plots=residualpanel;
Class id plate lot;
Model Dmm= id;
Random lot plate(lot) id(plate lot);
Title "Agreement about Diameter";
Title2 "Low serum";
Run;
To calculate my confidence intervals, I am using the suggestions from https://stats.idre.ucla.edu/sas/faq/how-can-i-bootstrap-estimates-in-sas/ with the following code:
proc sort data=lowserum;
by ID;
%let rep=500;
proc surveyselect data=lowserum out=bootsample
seed=1347 samprate=1 method = URS outhits rep=&rep;
cluster plate;
strata id;
run;
ods listing close;
Ods output covparms=T;
Proc mixed data= bootsample;
by replicate;
Class Id plate lot;
Model Dmm= id;
Random lot plate(lot) id(plate lot);
Title “Confidence intervals aroun Low Serum”;
Run;
proc transpose data=t out=t1 (rename=(Col1=Lot Col2=PlateLot Col3=LotPlateID Col4=Residual));
by replicate;
var Estimate;
run;
proc print;
run;
Data t2;
set t1;
Res= (Residual)/(Lot+PlateLot+LotPlateID+Residual);
ID= (LotPlateID)/(Lot+PlateLot+LotPlateID+Residual);
Plate= (PlateLot)/(Lot+PlateLot+LotPlateID+Residual);
Lots= (Lot)/(Lot+PlateLot+LotPlateID+Residual);
run;
proc print;
run;
However, my data has hierarchical clustering that is not being captured in my surveyselect procedure and my calculated confidence intervals do not contain my VPCs. I would like to have surveyselect return 21 observations per plate and for those observations to be equally distributed between 3 IDs.
Thank you!
