Josef Perktold

ENH: quantile parameterized distributions

4

I found this by chance https://en.wikipedia.org/wiki/Quantile-parameterized_distribution assumes bounded support of distribution It doesn't look popular enough as serious user facing distribution, but might be useful for some cases. I looked...

MAINT: failing style check: E721 do not compare types

1

example `if type(corr1) == np.ndarray:` should either use `is` or `isinstance`. Most likely we want `is` in those cases, i.e. rule out subclasses of ndarray (which might not have compatible...

parking some references, I did not look at them topic: outlier robust estimation of highly skewed distributions. related issues, e.g. #8690 #9148 #9146 #9128 Most standard theory for outlier robust...

ENH: make sure calling summary after remove data works if cached

1

see https://github.com/statsmodels/statsmodels/issues/9147#issuecomment-1939106367 currently `summary()` does not work for all models after remove_data has been called even when summary was called before remove_data unit tests for remove_data currently only check that...

- [ ] RobGLM: M-estimator explicit for logit, Poisson and Gamma - explicit formulas for expectation exist for those 3 distributions. good summary and formulas in Heritier et al ch.5...

ENH: right truncated count models for oulier trimming

1

related to robust regression, e.g. #3315 , https://github.com/statsmodels/statsmodels/issues/8690#issuecomment-1913715516 problem is imposing fisher consistency. one possibility would be to have a full MLE model that is right truncated at the trimming...

ENH Roadmap outlier robust covariance or scatter matrix

4

see - #3221 - #8129 - #9227 **update** basic models are merged, extensions missing, e.g. essentially only biweight norm has full support Maronna and Yohai 2017 recommend MM-estimator for p...

ENH: Outlier robust covariance - rebased

29

rebased version of #3230 (merge conflicts mainly in imports, not too bad) overview ### in `robust.covariance`: `cov_gk` GK (no reference in docstring) `cov_ogk` OGK based on Maronna and Zamar 2002...

ENH: maximum trimmed likelihood

3

I mentioned it in several issues but don't see a specific issue for it see also #3273 and #3284 This is the MLE version of least trimmed squares LTS or...

ENH: combined model checks, diagnostics, outlier influence, multicollinearity

1

support semi-automatic model checking, diagnostics, data problems, ... #6381, #1908 sanity checks, for model/data (pre-estimation checks), not post-estimation results. #1126 White's specification test #7834 diagnostic classes, get_diagnostic " Even with...