p-value and confident intervals with logistic regression?

Description

Steps/Code to Reproduce

Expected Results

Actual Results

Versions

Hi, Could it be possible to get p-value and confident intervals with logistic regression? If not, how could I get them? I tried with Logit in statsmodel, but it always output NAN value for coefficient and p-values. Thanks a lot!

[achievermina]

Can I work on this?

[rth]

See https://datascience.stackexchange.com/questions/15398/how-to-get-p-value-and-confident-interval-in-logisticregression-with-sklearn

There was also some an earlier discussion about this https://github.com/scikit-learn/scikit-learn/issues/6773

Although computing these for all models may be out of scope, I still think that it would be useful to add a small section on confidence intervals to the docs with pointers of where to look: e.g.

• statsmodels
• maybe an example of boostrapping with a simple linear model
• listing models that can be used to compute the confidence interval (e.g. Gradient Boosting, Gaussian Process Regression, random forest with http://contrib.scikit-learn.org/forest-confidence-interval/
• for models that have predict_proba mentioning the difference between that information and e.g. confidence intervals computed by boostrapping