JISCMail - FSL Archives

Firstly, the t-tests are directional, e.g. 1 -1 looks for A>B and -1 1 looks for B>A. In contradistinction, the f-tests test the power explained by the contrasts, and so lose directional information, e.g. an f-test containing 1 -1 or an f-test containing -1 1 are identical and both look for where either A>B or B>A.

Because t-tests in Feat are directional, they are one-tailed tests (we only look for significance in the positive tail of the null distribution). However, as f-tests are non-directional, an f-test containing a single t-contrast is only equivalent to a _two-tailed_ t-test on that contrast (where we look for significance in both tails of the null distribution).

Hence, you would need to double your p-value for the f-test in order to get the same probability under the tails, i.e. for positive and negative t-contrasts thresholded at p=0.05 you need to use a p=0.1 for the f-contrast to capture all of the former.

In short, your A-B zfstat map and B-A zfstat map will be identical (as f-tests are non-directional). And both should be the same as the combination of the A-B and B-A maps if you doubled the p-values used for the t-tests when you do the f-contrast. Without that adjustment, voxels could be passing one of the t-tests but not the f-tests - as appears to be happening in your data.

On 10 Jan 2007, at 17:40, Michael wrote:

Hi, it looks like you're missing the subject-specific mean modelling
- your EVs are all empty from EV2 onwards - see the example in the
manual.
Cheers.

Oh no, I just didn't add my subject-specific mean modelling in message that
I posted.

Here is my example: 10 subjects 2 conditions, paired-test.

Group EV1 EV2 EV3 EV4 EV5 EV6 EV7 EV8 EV9 EV10 EV11
1 1 1 0 0 0 0 0 0 0 0 0
1 1 0 1 0 0 0 0 0 0 0 0
1 1 0 0 1 0 0 0 0 0 0 0
1 1 0 0 0 1 0 0 0 0 0 0
1 1 0 0 0 0 1 0 0 0 0 0
1 1 0 0 0 0 0 1 0 0 0 0
1 1 0 0 0 0 0 0 1 0 0 0
1 1 0 0 0 0 0 0 0 1 0 0
1 1 0 0 0 0 0 0 0 0 1 0
1 1 0 0 0 0 0 0 0 0 0 1

1 -1 1 0 0 0 0 0 0 0 0 0
1 -1 0 1 0 0 0 0 0 0 0 0
1 -1 0 0 1 0 0 0 0 0 0 0
1 -1 0 0 0 1 0 0 0 0 0 0
1 -1 0 0 0 0 1 0 0 0 0 0
1 -1 0 0 0 0 0 1 0 0 0 0
1 -1 0 0 0 0 0 0 1 0 0 0
1 -1 0 0 0 0 0 0 0 1 0 0
1 -1 0 0 0 0 0 0 0 0 1 0
1 -1 0 0 0 0 0 0 0 0 0 1

And then my contrasts are:

title EV1 EV2 EV3 EV4 EV5 EV6 EV7 EV8 EV9 EV10 EV11
A-B 1 0 0 0 0 0 0 0 0 0 0
B-A -1 0 0 0 0 0 0 0 0 0 0

Shouldn't this be correct, no?

Also I want to add F-tests to my contrasts because I want to detect
activation for both postive and negative responses.

So my inital setup wiht contrasts was:

title EV1 EV2 EV3 EV4 EV5 EV6 EV7 EV8 EV9 EV10 EV11 F1 F2
A-B 1 0 0 0 0 0 0 0 0 0 0 * 0
B-A -1 0 0 0 0 0 0 0 0 0 0 0 *

But something strange is happening.

Is there something else I'm missing. Because when I look at my activation
maps:

A-B zstat contrast map - I see activation in areas I expect

B-A zstat contrast map - I see nothing at all

A-B zfstat contrast map - I see nothing at all

B-A zfstat contrast map - I see nothing at all

I just think there is something I'm not doing right.

Your help would greatly be appreciated.

Thanks