predict_confidence #77
Conversation
Handle prediction that returns a confidence interval
| Models that drop the intercept will be fitted without one: the intercept term will be | ||
| removed even if explicitly provided by the user. Categorical variables will be expanded | ||
| Models that drop the intercept will be fitted without one: the intercept term will be | ||
| removed even if explicitly provided by the user. Categorical variables will be expanded | ||
| in the rank-reduced form (contrasts for `n` levels will only produce `n-1` columns). |
There was a problem hiding this comment.
Everything above here is just my text editor removing trailing whitespace
Codecov Report
@@ Coverage Diff @@
## master #77 +/- ##
==========================================
- Coverage 93.03% 72.8% -20.23%
==========================================
Files 6 6
Lines 287 364 +77
==========================================
- Hits 267 265 -2
- Misses 20 99 +79
Continue to review full report at Codecov.
|
|
Is this ready to go in, or do you need anything more from me? |
|
That's fine with me (I've approved!). |
|
Thanks @nalimilan . @andreasnoack just remarked on Slack that he thinks the return type here should be a DataFrame not a 3-column Matrix. Do you agree with this? I can make a new PR. |
|
Maybe I didn't read the conversation carefully enough but it did seem to me as if people thought a table was appropriate here. |
|
This was merged before that discussion took place |
|
@mkborregaard Would you mind that this is reverted while you prepare the new PR? I'd like to get the GLM docs set up again but that requires a new release of StatsModels. |
|
That's fine I expect to fix this tomorrow |
This reverts commit a64cb50.
Revert "Handle prediction that returns a confidence interval (#77)"
Handle prediction that returns a confidence interval
cc @nalimilan
cf JuliaStats/GLM.jl#253