Yes the zstat image has both signs in it, but the thresholding and
the statistical tests are only done on the positive side.
If you just want to visualise +ve and -ve you can get away with one
contrast, and fslview, but if you want to do the statistical tests on
each, you need separate contrasts for +ve and -ve
T
On 10 Jul 2007, at 04:55, Vinod Venkatraman wrote:
> Hi Mark,
>
> I think the motivation of the question was to understand how FSL
> treats
> deactivations. In some other packages, A-B shows regions where A >
> B in one
> color and regions where B > A in another color. So, its redundant
> in a sense
> to do the two contrasts separately. But definitely that's not the
> case with
> FSL. But I am still wondering if the zstat image still has both +
> and -
> values in it, where + values indicate regions where A > B and -ve
> values
> indicate regions where B > A?
>
> Thanks
> Vinod
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf
> Of Mark Jenkinson
> Sent: Monday, July 09, 2007 7:28 PM
> To: [log in to unmask]
> Subject: Re: [FSL] generation of contrast maps
>
> Hi,
>
> Each of the contrasts does a one-tailed t-test and so in A-B you
> only see
> areas where the response to A is significantly greater than the
> response to
> B, while in B-A it is the other way around.
>
> In general you want to put in both contrasts in order to understand
> more
> about how the two conditions relate. The maps produced are
> normally very
> different indeed, and often one of them is blank (or almost so) if one
> condition engages a subset of the regions (and less strongly) than the
> other.
>
> All the best,
> Mark
>
>
>
> On 9 Jul 2007, at 22:45, Dharol Tankersley wrote:
>
>> Dear Listserv:
>>
>> I am trying to understand how FSL is generating the contrast map
>> A- B,
>> and how this is different from the contrast map B-A. What happens if
>> there is a deactivation in one of the conditions? The two maps do not
>> seem to be precise opposites of one another, and even if
>> deactivations
>> in some voxels are rounded to zero, this doesn't seem sufficient to
>> explain the large differences we are seeing in the A-B and B-A
>> contrast maps being generated.
>>
>> Thank you,
>>
>> Dharol
>>
>>
>> Dharol Tankersley
>> Dr. Huettel's Laboratory
>> Brain Imaging and Analysis Center
>> Duke University
>> 919-681-9935
>>
>>
|