Moving OLS in pandas

[Michael S]

Is movingOLS being moved from pandas to statsmodels?  It looks like the documentation is gone from the pandas 0.13.0.  Its an easy enough function to roll my own rolling window around statsmodel functions, but I always prefer more tested functionality.

(It used to be here: http://pandas.pydata.org/pandas-docs/version/0.10.0/computation.html#standard-ols-regression with the message "We’ll leave it to the user to explore the docstrings and source, especially as we’ll be moving this code into statsmodels in the near future.")

Thanks as always for all of the hard work.

Michael

[Skipper Seabold]

I can't speak to the documentation, but the plan has always been to
move the modeling code from pandas to statsmodels. A PR would be very
welcome, if even to get the ball rolling.

Skipper

[Michael S]

Thanks for the explanation, I will aim for a pull request later in the month.  
I appreciate it.

[josef...@gmail.com]

On Wed, Dec 4, 2013 at 2:46 PM, Michael S <michael....@gmail.com> wrote:
> Thanks for the explanation, I will aim for a pull request later in the
> month.
> I appreciate it.

Does anybody care about moving the git history (git filter-branch), or
will we move only the code?

Josef

[Skipper Seabold]

On Wed, Dec 4, 2013 at 8:09 PM, <josef...@gmail.com> wrote:
> On Wed, Dec 4, 2013 at 2:46 PM, Michael S <michael....@gmail.com> wrote:
>> Thanks for the explanation, I will aim for a pull request later in the
>> month.
>> I appreciate it.
>
> Does anybody care about moving the git history (git filter-branch), or
> will we move only the code?

I don't. The code that I've looked at would require a refactor anyway.

Skipper

[John Marino]

I'm trying to move my analytical work from SAS to the python ecosystem.  I'm a n00b, but finding it remarkably easy to replicate/verify my analyses.  One of my SAS scripts produces and stores daily moving-window regressions for all US exchange-listed stocks vs. the overall market, going back 10 years.  Just yesterday I wrote a python/pandas script that got exactly the same results (12k+ stocks, 19M+ output observations) using the MovingOLS object in pandas.  Jeff Reback tells me it is no longer supported in pandas.  Are you all planning to include it officially in statsmodels?  Like the OP, I'd rather not own it locally, but have it part of the tested framework.

[Skipper Seabold]

On Thu, Mar 5, 2015 at 7:57 AM, John Marino <jdma...@alumni.princeton.edu> wrote:

I'm trying to move my analytical work from SAS to the python ecosystem.  I'm a n00b, but finding it remarkably easy to replicate/verify my analyses.  One of my SAS scripts produces and stores daily moving-window regressions for all US exchange-listed stocks vs. the overall market, going back 10 years.  Just yesterday I wrote a python/pandas script that got exactly the same results (12k+ stocks, 19M+ output observations) using the MovingOLS object in pandas.  Jeff Reback tells me it is no longer supported in pandas.  Are you all planning to include it officially in statsmodels?  Like the OP, I'd rather not own it locally, but have it part of the tested framework.

If you're willing to put in the PR instead of having it local, that would be a great help in moving the support to statsmodels. We can help you along.

Skipper

[John Marino]

Ha!  I said I was a noob: I don't know what a PR is.  (I *think* it stands for "pull request", but I don't know what a "pull request" is.)  If this will really be helpful to the cause, I have plenty of colleagues who can help with a PR.  

[Skipper Seabold]

On Thu, Mar 5, 2015 at 8:49 AM, John Marino <jdma...@alumni.princeton.edu> wrote:

Ha!  I said I was a noob: I don't know what a PR is.  (I *think* it stands for "pull request", but I don't know what a "pull request" is.)  If this will really be helpful to the cause, I have plenty of colleagues who can help with a PR.  

Fair enough. I created an issue to keep track of this. I don't know how soon it will happen, but should be fairly straightforward to do. Any help is always welcome.

https://github.com/statsmodels/statsmodels/issues/2302