0

I am trying to run multi-level modeling using imputed dataset.

imp <- mice(data, m= 20, method = "pmm")
out1 <- with(data= imp, exp = lmer(issue.atti1 ~ age+gender+race+edu+pre1+pre2+group1+(1|state))

out1.pool <- pool(out1)
summary(out1.pool)

The summary of the pooled output provided fixed effects and intercepts, but not ICC (intra-class correlation) and AIC.

I tried this way:

sapply(out1.pool $analyses, AIC)

But it provides the list of AIC not AIC of pooled results.

#[1] 18229.26 18156.71 18197.92 18180.91 18155.70 18180.63 18190.48 18268.03 18174.94 18131.65
#[11] 18174.40 18172.52 18200.76 18150.48 18156.84 18237.29 18195.19 18241.88 18225.97 18127.35

How could I get the pooled AIC? I don't think we should report all of AIC in the paper.

Also, I have no idea how to get ICC for this pooled model. performance::icc() does not work here.

Does anyone have any solutions?

noone
  • 21
  • 1
  • This is probably as much a statistical question as a programming question. A near-duplicate: https://stackoverflow.com/questions/51815570/how-to-extract-aic-and-log-likelihood-from-pooled-glm?rq=1 – Ben Bolker Apr 21 '22 at 17:57
  • Can you point to a **statistical** definition of what you want to do? If so, someone here can help you with the programming. If you can't, then this should go to [CrossValidated](https://stats.stackexchange.com) instead ... – Ben Bolker Apr 21 '22 at 18:50
  • This is not possible with `mice` as far as I know. `mean(sapply(out1.pool $analyses, AIC))` might do the trick. – Robert Long Apr 21 '22 at 21:13

0 Answers0