[pymvpa] Suspicious results

[Michael Hanke]

On Mon, Feb 28, 2011 at 05:58:00PM +0100, Nynke van der Laan wrote:
> Summary per label across chunks
>   label mean std min max #chunks
>    1      1   0   1   1     36
>    2      1   0   1   1     36
> 
> So, each chunk has the labels balanced (which means chance level is 0.5).
> 
> To mention: all data stems from 1 scanning session (approximately 15
> minutes). Each chunk is one trial (approx 16 sec in total).

Hmm, if each trial is in its own chunk and you do leave-one-chunk-out-CV:
that makes the training dataset in each fold disbalanced (since you take
only one trial of one label out of a balanced dataset), right?

Why not group two trials into one chunks -- one of each kind.

Michael


-- 
Michael Hanke
http://mih.voxindeserto.de