Josef Perktold
Josef Perktold
(just a semi-random thought) AFAIR, our t-test, oneway tests and similar compute variance using estimated means and not means estimated under the null hypothesis. e.g Welsh anova allows for heteroscedasticity...
(mainly parking a reference for an old idea) variance and covariance estimates are not very good in small or very small samples. One idea is to use penalized or shrinkage...
based on the comments in #2208 `risk ratio` is a useful effects measure that is often used in medical, epidemiological studies, and I guess other areas. It makes a good...
work in progress, I opened the PR so I don't loose track of this This is currently just a RobustNorm class. no unit tests yet, and might not have correct...
I don't have a good overview what the current status is and the roadmap for it should be. We have several open issues and prs, enhancement requests, code that needs...
based on discussion #8380 - multinomial Logit can be extended to counts similar to GLM-Binomial versus Logit (we can reuse or copy-paste large parts of code for MNLogit) - other...
https://stackoverflow.com/questions/50070927/anova-table-and-pairwise-test tukey-hsd based on mean and var/std as sufficient statistics. I have not looked at it, but this should be made possible and easy to add, for one way anova...
(post-hoc) multiple comparisons like all pairs comparisons need p-value corrections. We don't have any closed testing procedure that However, sandbox.stats.multicomp has several functions or classes for subsets or partitions of...
I would like to have a diagnostic test, hopefully based on residuals like LM/score tests, for cluster correlation, i.e. specification test equivalent of cluster robust standard errors. There is a...
effect size again related to #5896 I'm still trying to figure out which and how to add effect sizes Problem is that we might want to use different versions of...