[pymvpa] correlation between 2 matrices

[David Soto]

apparently there seems to be a way of using in-house pymvpa tools to do representational similarity of spatio-temporal events 

(i) segment my timeseries data into the relevant spatio-temporal events with 'extract_boxcar_event_samples' 

(ii) flat the space-time matrices to plain vectors (as reccomended before in this post) 

(iii) set these to 'chunks' for the subsequent RSA analyis 

(iii) run 'rsa.PDist' on those vectors 

this approach would tell me the extent to which my ROI is modulated by spatio-temporal features of my stimulus 

does this sonds right? 

thanks! 

david 

From: "David Soto" <d.soto at bcbl.eu> 
To: "Development and support of PyMVPA" <pkg-exppsy-pymvpa at lists.alioth.debian.org> 
Sent: Friday, October 28, 2016 11:24:07 PM 
Subject: Re: [pymvpa] correlation between 2 matrices 

thanks that is useful, however to clarify I meant it to be in the context of a representational similarity analyses 
so would this solution be similar option 7 in here ? http://www.pymvpa.org/generated/mvpa2.measures.rsa.pdist.html 
how about if I wanted to use a euclidean distance? not sure it could handle 2 spatio-temporal matrices 
best 
david 


From: "Brian Murphy" <brian.murphy at qub.ac.uk> 
To: "Development and support of PyMVPA" <pkg-exppsy-pymvpa at lists.alioth.debian.org> 
Sent: Friday, October 28, 2016 9:58:45 PM 
Subject: Re: [pymvpa] correlation between 2 matrices 



If I understand your query correctly, I think numpy could do it, first by flattening the matrices to simple vectors, and then calculating the correlation between those, e.g. 

np.corrcoef(np.array(matrix1).flatten(), np.array(matrix2).flatten()) 

... assuming you're using the matrix datatype. If they are vanilla 2D arrays, then you shouldn't need to recast: 


np.corrcoef(matrix1.flatten(), matrix2.flatten()) 

best, 

Brian 


On 28/10/16 20:44, David Soto wrote: 



hi, 

I wonder whether PyMVPA can compute a similarity index (correlation) between two 2D matrices, each comprised of 12 seconds & 300 voxels? 

thanks! 
david 


_______________________________________________
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 




_______________________________________________ 
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 

_______________________________________________ 
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 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20161030/9964e69e/attachment.html>