search

statsmodels / statsmodels

Statsmodels: statistical modeling and econometrics in Python
http://www.statsmodels.org/devel/
BSD 3-Clause "New" or "Revised" License
9.98k stars 2.87k forks source link

ENH: confidence interval for Logit, seaborn #2508

Open josef-pkt opened 9 years ago

josef-pkt commented 9 years ago

I haven't looked at the details. This is mainly a reference to a (down stream) use case.

seaborn has bootstrap confidence intervals for logistic regression, which is slow. We should have them available without bootstrap (via endpoint transformation of linear prediction confidence intervals)

https://github.com/mwaskom/seaborn/issues/347#issuecomment-60974777 http://tomwallis.info/2014/10/29/high-level-plotting-in-python/

josef-pkt commented 6 years ago

https://stackoverflow.com/questions/47414842/confidence-interval-of-probability-prediction-from-logistic-regression-statsmode

GLM now has get_prediction, but there is no version for discrete. Generic version would require link function, but model group specific would be ok, e.g. binary models define cdf as mean or inverse-link function, simple count models use log-link. (new multipart count models are more difficult, and have more predict which)

josef-pkt commented 2 years ago

get_prediction for Logit is included in #7166