Phillip Alday
Phillip Alday
I like the idea of being able to set contrasts within a formula, but that's different from just vaguely marking something as categorical, which doesn't actually achieve the ostensible goal...
This is diverging into one of my rants, but > As an RA working on larger projects, I don't always have the chance to control the way the data is...
All or nothing. Either you're willing to accept that a formula is interpreted in the context of its data and the code above it or you want the formula to...
> If the reason we shouldn't add this is because the contrast information isn't explicitly available then we should probably not assume symbols or strings are categorical unless the data...
> Why couldn't you just do the same with other variables that are declared as "categorical", assume dummy coding unless told otherwise? This is really what I'm getting at. If...
Yeah, that's about as far as I had gotten as well when I ran out of steam late local time. `deepcopy(::Missing)` is defined, and given that `missing` is a singleton,...
Ignore the horrible hacking of `getindex` -- that will not stay.
This would actually fix a number of current issues with handling missings, especially when expanding categorical variables into contrasts.
@matthieugomez That's not too surprising given the underlying issue (and how it expands out), but thanks for the example.
Why MixedModels.jl defines its own LRT: 1. We needed it while the `lrtest` PR was stalled and so we made it and released it. 2. `isnested` is a challenging question...