[Pkg-exppsy-pymvpa] New splitter?

[Michael Hanke]

Hey,

On Mon, Apr 21, 2008 at 11:59:14AM -0400, Per B. Sederberg wrote:
> Howdy Michael:
> 
> I having a little trouble with the new splitter.  I think everything
> is there to do what I'd like, but I can't figure it out based on the
> Splitter code.
> 
> I have a custom dataset attribute called tasks and each TR's task is
> set to be 0,1, or 2 depending on what task the participant was
> performing at that time.  I'd like to return a single split of the
> data, training on the 0 and 1 tasks and testing on the 2 task.  I'd
> also like to ensure that the same number of each task are in the
> training dataset, but I don't care about the testing dataset.
> 
> I think I have almost what I want with the following custom splitter:
> 
> CustomSplitter([(None,[2])],nperlabel='equal')
> 
> However, it looks like the nperlabel is getting applied to both the
> training and the testing samples.  The old splitter allowed you to
> apply this separately for the first and second samples via
> nfirstsamples and nsecondsamples, but it looks like this functionality
> was removed with the latest other improvements, which are totally
> awesome, BTW.
Glad you like them ;-)

This should solve your problem:

	CustomSplitter([(None,[2])],nperlabel=['equal', 'all'])


Does it work?

Cheers,

Michael


-- 
GPG key:  1024D/3144BE0F Michael Hanke
http://apsy.gse.uni-magdeburg.de/hanke
ICQ: 48230050