Hi Raj,
I am not sure about this, but your problem sounds like you need a
multivariate analysis. Correlating all your behavioral measures with all
your fMRI activation from different ROIs could be hard to interpret,
especially in the case of correlations among both sets of variables (the
behavioral measures and the activation). In that case different
correlation coefficients could potentially explain the same variance in
the data and you would not be able to interpret them independently.
Again, I am not certain about this, but it seems to me that you might be
able to use a canonical correlation analysis for your problem. This
analysis is suited to deal with a multiple variables on both side of the
equation (e.g. multiple dependent and independent variables).
Cheers,
Jan
Rajeev Raizada wrote:
> On Thu, 20 Apr 2006, Thilo Kellermann wrote:
>
>
>> Hi Raj,
>>
>> thanks a lot for sharing your ideas about the permutation tests. Maybe
>> you are also interested in an SAS script that has been written by Warren
>> Bilker for testing correlated correlations, which is also based on
>> permutation tests.
>> This can be found at:
>> http://www.cceb.upenn.edu/main/people/bilker.html
>>
>
> Many thanks indeed for the info.
>
> I had a look at Prof. Bilker's code, although I haven't been
> able to run it yet because I don't have SAS.
>
> It looks infinitely more sophisticated than the simple little
> Matlab script that I posted, and it appears to cover a far broader
> range of different situations.
>
> I believe, but am not certain, that for the simple situation
> of when you correlate a bunch of behavioural scores with a bunch
> of ROI-activities, then the basic recipe should suffice
> of shuffling the subjects, calculating the correlations,
> appending them to the cumulative pile of results, repeating,
> and then looking at the distribution of the end result.
>
> That's just five lines of Matlab in the main while-loop,
> if you ignore my overly-verbose comment lines.
> http://faculty.washington.edu/raizada/raj_permutation_test_behav_fMRI.m
>
> My question is:
> Is this simple approach statistically valid?
> Or do I need to use a much more sophisticated set of analyses
> like the stuff that Prof. Bilker's SAS code implements?
>
> Raj
>
>
--
Dr. rer. nat. Jan Gläscher Caltech Brain Imaging Center
+1 (626) 395-4976 (office) Caltech, Broad Center, M/C 114-96
+1 (626) 395-2000 (fax) 1200 California Blvd
[log in to unmask] Pasadena, CA 91125
|