I used MixedModels package a couple of times in the past as a faster alternative to R’s lmer. Recently I downloaded latest official Julia version (0.6.3) and tried with some datasets. I compared the results with old MixedModels and Julia version, and I found that random effects are not matching at all (using ranef(lmeModel, named=true)). Digging little deeper i managed to match them using the following snippet:
randomEff = transpose(ranef(lmeModel, named=true))
rows, columns = names(randomEff)
randomDF = DataFrame(randomEff)
names!(randomEff.colindex, map(parse, columns))
randomEff[:store] = unique(lmeModel.trms.f)
dfCols = randomEff.colindex.names
randomEff = stack(randomEff,dfCols[1:end-1], :store)
Now random effects appear very close to old results.
The question is: Am I doing something wrong or there is an issue with random effects in new version?
Please note that I have very little experience with Julia.