topic Re: Bootstrapping a CI or Std Error in a Mixed Model in Statistical Procedures

Bootstrapping a CI or Std Error in a Mixed Model

Daisy2 — Thu, 06 Feb 2020 21:39:59 GMT

I'm using a mixed model with multiple random and fixed effects (REML) and continuous y and I would like to try using bootstrapping to get more accurate CI or SE values since even with transformation or BoxCox my y-data are not always normal. I don't find a proc to do this. Can someone help? I am quite familiar with JMP 14 and use SAS 7.154 a little bit. Thanks. Here's the code I'm starting with:

proc mixed data=phenoall;
  class hyb n pd loc;
  model nce=hyb n pd n*hyb pd*hyb n*pd;
  random loc block(loc) pass(loc) loc*n loc*hyb;
  lsmeans hyb / adjust=tukey;
run;

Re: Bootstrapping a CI or Std Error in a Mixed Model

PaigeMiller — Thu, 06 Feb 2020 23:28:53 GMT

Everything you ever wanted to know about bootstrapping in SAS

https://blogs.sas.com/content/iml/2018/12/12/essential-guide-bootstrapping-sas.html

Re: Bootstrapping a CI or Std Error in a Mixed Model

Daisy2 — Fri, 21 Feb 2020 23:30:13 GMT

Thanks. Just curious, does the bootstrapping give you better SE for the fixed effects or the random effects?