[pymvpa] Pattern localization
Yaroslav Halchenko
debian at onerussian.com
Fri Apr 24 17:26:25 UTC 2009
> from the source code:
> "WARNING: Highly experimental/slow/etc: no theoretical grounds have been
> presented in any paper, nor proven"
>
> Sounds like: If you use it (right now), you must be total idiot ;-)
or 'researcher, genius, etc' ? ;) it is just based on observation we
had, which is partially supported with a theory and not-yet been
explored in details anywhere we looked at... so, if you are up for some
collaboration -- we could talk about it in detail ;)
> But, would be very nice to have some p-values and map it with a
> threshold using a multiple comparison correction criterion. I'm really
> looking forward to this transformation.
well... there is no best way imho to correct for multiple comparisons...
Bonferonni is too concervative, FDR might be the one close to the best I
guess. Non-parametric p-values (and transformed under assumption of
Normal distribution z-scores) could be obtained from permutation
testing. That 'idiot' beast above might be the closest best shout for
this particular case (sensitivities) due to one nice property of rdist
;)
> OK! I assume with "baseline condition" you mean the explicit baseline
> ("rest") and not the implicit baseline which would be everything else,
> except FACE and HOUSE (like e.g. FSL does), right?
rright -- I believe that is what we have used.
>> SVM assigns weights per each category -- positive for +1, negative for
>> -1. So it is just to sum up corresponding SVs accordingly... let me
>> simply actually patch libsvm's LinearSVMWeights for now... I will let
>> you know whenever it is done
> Oh great! Thank you very much for putting so much work into this!
working on it... although on the way stumbled upon some other elderly
code which needed some attention ;)
--
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student Ph.D. @ CS Dept. NJIT
Office: (973) 353-1412 | FWD: 82823 | Fax: (973) 353-1171
101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW: http://www.linkedin.com/in/yarik
More information about the Pkg-ExpPsy-PyMVPA
mailing list