broom icon indicating copy to clipboard operation
broom copied to clipboard

Added Wald confidence intervals for glm - resolves #989

Open lcgodoy opened this issue 3 years ago • 5 comments

For this pull request, the feature requested in issue #989 was included in the package. The conf.method parameter was included in the tidy.glm method. Its possible values are, the default, "plik" for computing the profile likelihood CI or "wald" for the Wald CI. In order the make it work consistently, the function broom_confint_terms had to be modified. These changes have not impacted other methods that depend on broom_confint_terms.

lcgodoy avatar Oct 24 '21 15:10 lcgodoy

Hey there, thanks for the PR!

Apologies on the change of mind here, but I no longer feel that moving forward with this feature is the right move. Given where this package is at in its lifecycle and my own (in)ability to maintain this functionality (and the addition in scope it otherwise welcomes) well, I don't feel that I'm comfortable with adding features to the bread-and-butter stats tidiers.

simonpcouch avatar Dec 28 '21 14:12 simonpcouch

@simonpcouch I'd be willing to take on maintenance duties for this if that changes your opinion here. This change really should only be local to the tidy.glm() and we could implement it without touching broom_confint_terms at all, which should keep the maintenance burden low and avoid spooky behavior.

alexpghayes avatar Dec 28 '21 16:12 alexpghayes

@alexpghayes I'm for it! Leaving broom_confint_terms alone sounds like the right move to me as well.

simonpcouch avatar Dec 28 '21 16:12 simonpcouch

@lcgodoy I'll add a quick code review momentarily.

alexpghayes avatar Dec 28 '21 16:12 alexpghayes

Hi everyone,

I just added the issue https://github.com/tidymodels/broom/issues/1074 which is highly related to this PR.

There, I suggest that when this feature is accepted, it should become the default behaviour of tidy.glm().

This might lead to minor breaking changes but it would correct the mismatch between p-values and confidence intervals, which I find important.

DanChaltiel avatar Jan 12 '22 10:01 DanChaltiel

Hi, I'm also very interested in the ability of specifying confidence interval method. What's the status of this request? Can we see it implemented soon?

snhansen avatar Feb 07 '23 09:02 snhansen

As we've been unable to bring together the development + maintenance energy for this changes in the last two years, I'm going to go ahead and close. Thanks all for the efforts.🌸

simonpcouch avatar Jul 26 '23 21:07 simonpcouch

This pull request has been automatically locked. If you believe the issue addressed here persists, please file a new PR (with a reprex: https://reprex.tidyverse.org) and link to this one.

github-actions[bot] avatar Aug 10 '23 00:08 github-actions[bot]