[pymvpa] Error when using pearsonr with ridgeregression

[Yaroslav Halchenko]

Hi Thomas,


On Thu, 15 Jan 2015, Thomas Nickson wrote:

>  /home/orkney_01/tnickson/Programming/pyVirtualEnv/lib/python2.6/site-packages/mvpa2/base/dataset.pyc in __init__(self, samples, sa, fa, a)
>      213         if not hasattr(samples, 'dtype'):
>      214             raise ValueError(
>  --> 215                 "AttrDataset only supports dtypes as samples that have a "
>      216                 "`dtype` attribute that behaves similar to the one of an "
>      217                 "array-like.")

>  ValueError: AttrDataset only supports dtypes as samples that have a `dtype` attribute that behaves similar to the one of an array-like.

good way to troubleshoot such troubles when working in ipython, is to
enable pdb (just type %pdb) so it would fall into debugger right at that
point.  Then you could inspect

(Pdb) print samples
(0.26726124191242479, 0.010880186798833041)

to get a hint that pearsonr returns a tuple of both correlation and p value.
It still smells like a bug in PyMVPA that we don't wrap it up properly but as
an overcome you would just need to choose which one to use, e.g.

errorfx=lambda x,y: pearsonr(x,y)[0],

But then let's get back altogether to your analysis -- you are using a
classifier, so dealing with classes and then using correlation coefficient as a
measure (for ordered data) -- doesn't sound quite right unless there is some
agenda behind your dataset's targets ;)


-- 
Yaroslav O. Halchenko, Ph.D.
http://neuro.debian.net http://www.pymvpa.org http://www.fail2ban.org
Research Scientist,            Psychological and Brain Sciences Dept.
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834                       Fax: +1 (603) 646-1419
WWW:   http://www.linkedin.com/in/yarik