BUG/ REF conf_int and use_t #1812

Closed
josef-pkt opened this Issue Jul 8, 2014 · 1 comment

Projects

None yet

1 participant

@josef-pkt
Member

I thought every inferential statistic uses now use_t. However, while cleaning up the docs I saw that LikelihoodModelResults.conf_int has normal distribution hardcoded.

I thought I have generic tests for this. But maybe only for the defaults.

relevant PR is #1189
It didn't change conf_int which is a mistake, but it has unit tests for conf_int. So I guess there is no test case with non-default conf_int.

linear_model.RegressionResults.conf_int uses use_t
code comment: should be obsolete except super LikelihoodModelResults.conf_int doesn't use use_t

reading through the comments in PR1189, I see it was only/mainly for regression, not for all models.
need another round of refactoring to make the missing changes in base and other models.

@josef-pkt
Member

GEEResults also defines its own conf_int hardcoded for normal distribution

@josef-pkt josef-pkt added this to the 0.6 milestone Jul 16, 2014
@josef-pkt josef-pkt closed this in fc2ae46 Jul 18, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment