[pymvpa] Dataset with multidimensional feature vector per voxel

[Ulrike Kuhl]

Thanks again for your super quick and super helpful reply! 

There is still one thing that I don't quite get at the moment: 

As expected, 'dsall' is a numpy.ndarray of dimensions 1*(numberVoxels*numberSubjects*numberParameters) - so just a really long vector.
Since this is 'only' a vector, it does not have any attributes or other information associated with it. This information is still contained in 'dss', which is a list of size (numberSubjects*numberParameters) containing the respective datasets.
I don't see right now how to bridge this information to the 'dsall'-array. After all, 'sphere_searchlight()' should also be called with a dataset, using this dataset's coordinates to determine the local neighborhoods, right?
So I guess I still need the 'glue' to get it all together... ;-)
Can you give me a hint on that?

Cheers,
Ulrike

----- Original Message -----
From: "Yaroslav Halchenko" <debian at onerussian.com>
To: "pkg-exppsy-pymvpa" <pkg-exppsy-pymvpa at lists.alioth.debian.org>
Sent: Thursday, 5 November, 2015 15:14:24
Subject: Re: [pymvpa] Dataset with multidimensional feature vector per voxel

On Thu, 05 Nov 2015, Ulrike Kuhl wrote:

> Dear Yaroslav,

> thanks a lot for your reply.

> With your snipped it was really easy for me to set up the matrix as you described it. :-)
> For those interested, this is the code that works for me:


> sub_list = [list of subjects]
> param_list = [list of parameters]
> dss = []

> for sub_index, sub in enumerate(sub_list):
>     for suf_index, suf in enumerate(param_list):
>         ds = fmri_dataset('/path/to/image/file',mask='/path/to/mask/file')
>         ds.fa['modality'] = [suf] * ds.nfeatures             # for each feature, set the modality attribute
>         ds.fa['modality_index'] = [suf_index] * ds.nfeatures # this as numeric might come handy for searchlights later
>     if sub.startswith('L'):
>         learn = 1
>     elif sub.startswith('N'):
>         learn = 0
>         ds.sa['targets'] = [learn]                           
>         ds.sa['chunks'] = [sub_index]
>         dss.append(ds)
> dsall = hstack(dss)

Here is my tune up with # yoh: comments

####################################
sub_list = [list of subjects]
param_list = [list of parameters]
dss = []

for sub_index, sub in enumerate(sub_list):
    for suf_index, suf in enumerate(param_list):
        ds = fmri_dataset('/path/to/image/file',mask='/path/to/mask/file')
        # yoh: no need to broadcast -- should do automagically
        ds.fa['modality'] = [suf]             # for each feature, set the modality attribute
        ds.fa['modality_index'] = [suf_index] # this as numeric might come handy for searchlights later
    if sub.startswith('L'):
        learn = 1
    elif sub.startswith('N'):
        learn = 0
    else:
        # yoh: make it explicit, otherwise you would assign previous learn to next one not L or N
        raise ValueError("Do not know what to do with %s" % sub)
    # yoh: seems it was incorrectly indented into elif
    ds.sa['targets'] = [learn]
    ds.sa['chunks'] = [sub_index]
    dss.append(ds)
dsall = hstack(dss)

####################################

cheers

-- 
Yaroslav O. Halchenko
Center for Open Neuroscience     http://centerforopenneuroscience.org
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        

_______________________________________________
Pkg-ExpPsy-PyMVPA mailing list
Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa
-- 
Max Planck Institute for Human Cognitive and Brain Sciences 
Department of Neuropsychology (A219) 
Stephanstraße 1a 
04103 Leipzig 

Phone: +49 (0) 341 9940 2625 
Mail: kuhl at cbs.mpg.de 
Internet: http://www.cbs.mpg.de/staff/kuhl-12160