release 0.6

[josef-pkt]

ticket to keep milestone 0.6 open

What should still go into 0.6? What needs to be fixed?

update release notes and docs

open PR or issues

• GEE
  • MAINT: Refactor GEE to put categorical models in subclasses #1694
• Mixed
  • BUG: Fixed xrange probem in MixedLM profile; fixed complex number problem in MixedLM Hessian #1695
• PHReg, Survival
  • Proportional hazards survival regression model (a.k.a. Cox model) #1312
  • survival2
• GMM
• Robust
  • takes too much time to review
• discrete
  • [ ]
• Nonparametric
  • several by Padarn
• Stats
• Graphics
• general, base
• MixedLM wrapper MixedLM does not have a wrapper #1990

I changed the milestone to 0.7 on several PRs

Bugs

• Added error handling to MultiComparison #1927 PR ready
• BUG: Fix missing initialization of "formatted_rows" #1929
• WIP/BUG: fixed ProbPlot class and tests to work with general distributions #1433 probplot ?
• pacf has incorrect standard errors for lag 0 #1969 pacf conf_int
• A small bug in GenericLikelihoodModelResults.bootstrap() #1434 bootstrap in GenericLikelihood
• BUG: fix test error in Optimizer._fit_newton. #1901 numpy compat
• ENH: Add api for covstruct. Clear __init__. Closes #1917. #1981 api for GEE covstruct
• GEE formula interface does not take offset/exposure #1991 GEE.from_formula does not handle exposure/offset

and a few more PRs left with milestone 0.6

---

items to check that they are in release notes

• sequence of PRs with bug fixes, refactorings in GLM, especially offset and exposure MAINT: GLM #1734

[rgbkrk]

Hey there! When are you all expecting to release 0.6?

We were hoping to include this in IPython's Docker images for the scipystack as part of this PR: jupyter/docker-notebook#9, but noticed that 0.5 does not build in Python 3.4 (while the master branch does).

/cc @takluyver

[jseabold]

We are discussing. Hope to cut a RC asap.

[josef-pkt]

I'd like to get in these two #1927 #1929 It will take me a few hours tonight or tomorrow

I'd like this one #1433 probplots, but I have no idea what the status is. (I don't know whether the current code doesn't have any bugs left, that should get resolved with #1433.) It will take a few hours writing examples in a notebook to see how well it works.

I think it's too late for almost all of the other PRs. (I decided a month ago not to work through those, if we want to release fast.)

There are a few other issues that should/could get in, maybe even between rc and release.
One is to raise an exception on object arrays, as a temporary fix until we have full dtype handling.

I'm waiting for 0.6 to be out of the way to do more merging of larger PRs, hoping for a quick 0.7

(unit root is unfortunate, and I'd like to get half way through #1849 but adfuller is too popular to change things around without being sure about backwards compatibility. I haven't looked at unitroot in more than a month.)

[josef-pkt]

another one: no specific issue open yet: MixedLM is missing the ResultsWrapper.

[Edit: #1990]

[josef-pkt]

and smaller bugs like #1969 #1434

[josef-pkt]

another one #1901 numpy compatibility

[stonebig]

Hi josef-pkt,

Thanks for this awesome work. I would be glad if you could release 0.6 this week, as I can only get 0.5 at this point (as relying on Christoph Gohlke binaries).

[jseabold]

I'm going to finish checking the release docs and tag the release candidate today.

[jseabold]

The release candidate is tagged and available on PyPi. Please give it a spin.