Open benwhalley opened 7 years ago
Thanks. I'll investigate.
You could also take a look on a method from Ben Bolker, which would reduce dependencies on other packages: http://bbolker.github.io/mixedmodels-misc/glmmFAQ.html#predictions-andor-confidence-or-prediction-intervals-on-predictions
However, this approach does not take the uncertainty of random effect into account.
I wonder whether it would makesense to use
predictInterval
from merTools instead of the standardpredict
function? At presentprediction
returns NA for the fitted.se, but predictInterval would make this possible, and is also relatively fast:https://cran.r-project.org/web/packages/merTools/vignettes/Using_predictInterval.html
Perhaps merTools could be an optional dependency, and if not present we'd fall back to the standard predict method?