[pymvpa] Inspect every cross-validation fold

[Yaroslav Halchenko]

every confusion matrix (e.g. ca.stats for cross validation on classification)
stores each fold separately.  So you could get 1 per each fold from
.ca.stats.matrices.  Also there are .ca.stats.sets which store raw (targets,
predictions, estimates) per each provided split if you want to see at the
actual values.

also, in current master we extended confusion matrix printout with a new
entry in addition to plain # of sets:

# of sets            12       ACC(i) = 0.87-0.015*i p=0.3 r=-0.33 r^2=0.11

so if you had 12 folds from NFoldPartitioner(cvtype=1), this line would
tell either there was some linearly consistent trend in accuracies across
the folds.

hope this helps

On Wed, 06 Apr 2011, Roberto Guidotti wrote:

>    Dear all,
>    I'm trying to analyze a dataset with PyMVPA, but I want to know how the
>    classifier acts across folds in cross-validation or having a
>    ConfusionMatrix for every cv fold.
>    I've tried enabling in cv class 'training_stats' but the results is
>    incomprehesible for me!!
>    Is there a method or attribute or something else or the possibility to
>    implement this feature?
>    Thank you
>    Roberto

> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa


-- 
=------------------------------------------------------------------=
Keep in touch                                     www.onerussian.com
Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic