Implement StatsAPI.pvalue
#1719
Conversation
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #1719 +/- ##
==========================================
+ Coverage 85.82% 85.85% +0.03%
==========================================
Files 137 138 +1
Lines 8318 8337 +19
==========================================
+ Hits 7139 7158 +19
Misses 1179 1179
☔ View full report in Codecov by Sentry. |
| @@ -0,0 +1,26 @@ | |||
| using Distributions | |||
| using StatsAPI: pvalue | |||
There was a problem hiding this comment.
Would it make sense to have Distributions reexport pvalue? I'd be generally in favor but don't feel too strongly.
There was a problem hiding this comment.
Wouldn't this break HypothesisTests?
There was a problem hiding this comment.
Not after HypothesisTests extends pvalue from StatsAPI.
There was a problem hiding this comment.
Sure, but all existing releases of HypothesisTests would suddenly be broken, wouldn't they? Certainly, this could be fixed by modifying the registry but I think this makes it less attractive to re-export pvalue.
There was a problem hiding this comment.
I'll merge and release without re-exporting pvalue for now but this can be reconsidered and changed in a subsequent release if desired.
Co-authored-by: Alex Arslan <ararslan@comcast.net>
Currently
pvalue(::UnivariateDistribution, ...)is defined in HypothesisTests and owned by HypothesisTests. However, when switching HypothesisTests to StatsAPI 1.6 which definesStatsAPI.pvaluesuch definitions would be type piracy. Hence this PR adds them to Distributions.Note that StatsAPI is already an indirect dependency of Distributions via StatsBase, so this PR does not add packages to the dependency tree.
TODO: