Josef Perktold

Results 172 issues of Josef Perktold

some references (for the robust RLM case) http://scholar.google.com/scholar?cites=17895094206680664513&as_sdt=2005&sciodt=0,5&hl=en Question: if we impose a constraint and want to test inequalities, are we running into the parameter on the boundary of the...

wishlist
comp-regression
comp-stats

(inequality constraints in wald test with multiple parameter is a difficult topic, open issues for it) see #1382 and related for general discussion one computational detail Shapiro, Alexander. 1985. “Asymptotic...

type-enh
comp-base
comp-stats

We have issues for using Mallow's or Schweppe's weights but I don't see any functional form for it. We need references and functions for those. Then we could use var...

type-enh
comp-robust

related to #1379 more cov_types for RLM, M-estimators in general Arellano's slide referenced here https://github.com/statsmodels/statsmodels/issues/3335#issuecomment-268719056 http://www.cemfi.es/~arellano/survey-metrics-sep2014-tr.pdf use quantile regression as one of the examples. The section on cluster robust covariance...

type-enh
comp-regression
topic-covtype

followup to #696 specifically what inference can we use after estimation. question that I'm looking at right now is whether QuantReg should use t or normal distribution. (I'm adding a...

type-bug
design
comp-regression
topic-covtype

RLM has neither score, score_obs nor hessian Treating it as M/GMM estimator we need score_obs and hessian to compute generic sandwich cov_types. see #9190 and ... With score_obs and hessian...

type-enh
comp-robust
topic-covtype

I merged resistant estimators, S, MM for linear model. The starting sets use the most central exog observations (smallest Mahalanobis distance) These starting parameters will not work for nonlinear or...

type-enh
comp-robust

This should be a curated list of related packages, especially to statistics and econometrics topics that statsmodels does not (yet) provide. (not complete adding links as found, not sorted and...

comp-docs
FAQ

AFAICS, we don't have a t-distribution equivalent of the two functions `normal_power_het` and `normal_sample_size_one_tail` Also `ttest_power` might need a `ddof` keyword. However, I did not look right now how df...

type-enh
comp-stats

this should add stepwise regression, sweep algorithm and sequential OLS see #1809 currently it is mainly a script I wrote in 2012 converted to a module and two unit tests...

type-enh
comp-regression
comp-tools