statsmodels compatibility #17
Conversation
treat vectors of EventTimes like continuous terms. needs JuliaStats/StatsModels.jl#106 (to just copy continuous terms instead of converting to Float64).
|
You'll probably want to turn the StatsModels upper bound in the Project.toml into a lower bound in order for this to work, since |
|
I guess we'll need to wait until StatsModels is tagged, since otherwise the StatsModels version requirement is unsatisfiable. |
| ## StatsModels compatibility | ||
|
|
||
| StatsModels.concrete_term(t::Term, xs::AbstractVector{<:EventTime}, ::Nothing) = | ||
| StatsModels.ContinuousTerm(t.sym, first(xs), first(xs), first(xs), first(xs)) |
There was a problem hiding this comment.
Why not actually compute the mean, variance, min and max?
There was a problem hiding this comment.
Is that even defined for event times? How would the truncation be handled?
Maybe a better option is to add methods like mean(<:AbstractVector{<:EventTime})?
There was a problem hiding this comment.
Hard to tell... But it also sounds weird to pass identical values. Maybe nothing should be allowed?
|
Tests pass locally for me with StatsModels 0.6.0 but I've restarted the travis job to make sure... |
|
Has the StatsModels 0.6.0 tag not been merged to General or something? |
|
No it has but it lists julia >= 1 in the project.toml so that 0.7 build failed...do you care about 0.7 compatibility still? |
|
(I'll update the travis config and Project.toml in this PR if you want to drop 0.7 support...) |
|
Down with 0.7! Down with 0.7! |
|
Oh, I see, I was fooled because I was looking at the push build rather than the PR build. That's why I didn't see that it was only not working on 0.7. |
|
I guess also the version should be bumped right? |
|
Yep |
treat vectors of EventTimes like continuous terms. needs JuliaStats/StatsModels.jl#106 (to just copy continuous terms instead of converting to Float64).
fixes #10 , #16