[SciPy-User] Repeated Measure ANOVA

[Skipper Seabold]

On Wed, Nov 6, 2013 at 1:49 PM, Horea Christian <horea.christ at gmail.com>wrote:

> Hi guys, I would like to compare reaction times for a series of
> experimental conditions. My data comes from ~100 trial repetitions over 10
> participants (yielding ~1000 trials). I was told that just doing an ANOVA
> on this dataset would be improper, because the 1000 measurements are not
> truly independent - and that the proper way to do this is called a repeated
> measure ANOVA.
>
> I have tried to look for a scipy function for this and found nothing. In this
> relevant discussion<http://scipy-user.10969.n7.nabble.com/stats-repeated-measures-ANOVA-td15997.html>a participant pointed the following out:
>
> "Repeated measures" ANOVA is just a misnomer for using the "randomized
>> block design" as a substitute for not knowing MANOVA or Hotelling's T-
>> square test, and as such leads to conclusions that are very hard to
>> interpret. The real value of repeated measures ANOVA in medical
>> litterature is often to inform the reader that the authors don't
>> understand the statistics they use ;-)
>
>
> I would like to know whether I'm looking for the right thing at all, and
> if yes how I could accomplish this with scipy
>

Repeated Measures ANOVA is waiting a champion. I don't think it's going to
be entirely trivial to get it right, and I just don't have the bandwidth
right now to put in any (unpaid) time on this, though maybe it'll fall out
of our ongoing panel data work in statsmodels (it's unclear to me right
now).

https://github.com/statsmodels/statsmodels/issues/749
https://github.com/statsmodels/statsmodels/pull/786
https://github.com/statsmodels/statsmodels/issues/646

Skipper
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20131203/32a8fa04/attachment.html>