release 0.9 again

79 views
Skip to first unread message

josef...@gmail.com

unread,
Jun 13, 2017, 12:36:28 AM6/13/17
to pystatsmodels
(I'm leaving for a month in Europe in a bit more than a week, and my
computer and internet access is not clear yet.)


I would like that we release 0.9 in early fall, say September. We have
several compatibility fixes with pandas and numpy that we need to get
out.
Releasing after the summer leaves us a bit of time to review and fix
some issues.

I would still like to merge now a few PRs that might need post-merge
review and changes. Feedback, especially until a release, would be
very helpful and appreciated and would minimize post-release changes.

https://github.com/statsmodels/statsmodels/pull/3246
VECM and cointegration, should be fine with good test coverage. Some
names are not quite "obvious" but those are mostly internal

https://github.com/statsmodels/statsmodels/pull/3692
GLM var_weights and family cleanup. This looks good for parameter
estimation and wald inference, but our scale and dispersion handling
is still buggy for loglikelihood. Those are properties that are
currently not used, but need to be fixed.

https://github.com/statsmodels/statsmodels/pull/3303
rebased with a few changes in
https://github.com/statsmodels/statsmodels/pull/3663
Repeated Measures Anova for within effects with balanced data
looks good, maybe minor name changes could improve it

https://github.com/statsmodels/statsmodels/pull/3727
Generalized Poisson count model. This is the first model of one of our
GSOC project this year. It's almost ready to merge and will be needed
for additional models during the GSOC project.


Besides these, there are several smaller fixes, refactoring,
improvements, maintenance that will still be either necessary or
good-to-have for 0.9.

And there are optional merges for whatever is ready. (I have a few
open PRs where I don't want to wait much longer for the "big
inspiration" to get the perfect API if what we have is pretty good,
like segmented regression, comparing poison rates, some stats for
proportions, and similar.)

There are also other PRs that are 80% (+/- 10) finished in case
someone would like to take advantage of some opportunities for
reviewing and coding during summer (Is it called "Summer of Code" or
"Coding on the Beach"?)


Josef

Big Stone

unread,
Jun 18, 2017, 12:01:24 PM6/18/17
to pystatsmodels
Hi Joseph,

Many thanks for the update and timeline for statsmodels-0.9.

josef...@gmail.com

unread,
Jun 28, 2017, 1:52:25 PM6/28/17
to pystatsmodels
All four are now in master.

It's time to give it a spin, and provide feedback so we can fix some of the rough edges before the next release (as far as there are rough edges).
Documentation still has to catch up. (The game is figure out if this program is self explanatory without RTF(NE)M.)

Josef
Reply all
Reply to author
Forward
0 new messages