Matthew Whalley wrote:
> Sorry, I should have said that all the contrasts I'm using are F's -
> that's why with the imcalc method I specify two-tailed thresholds.
Ah, well, like I said, a wild guess. Also, I'm afraid, the only guess
I can come up at the moment... Hopefully someone else on the mailing
list can give you a better answer!
Sorry not to be of help,
Ged.
> If it makes it a little clearer, I'm looking for the activity common to
> three groups of patients (PTSD, Depressed & Control)., exclusively
> masked with the areas where one group has greater activity than another.
> I'm told that Rik Henson has said the imcalc way is valid, but was
> hoping that the quick & easy GUI way would give an equivalent result.
>
> Thanks for your help,
>
> Matthew
>
> On 12 Dec 2006, at 17:19, Ged Ridgway wrote:
>
>>
>> Hi Matthew,
>>
>> This is a completely wild guess, as I'm not sure I've understood
>> exactly what you're doing, but is it possible that method 1 is using
>> SPM's usual one-tailed t-contrasts, whereas in your second method you
>> specify two-tailed t thresholds?
>>
>> Ged.
>>
>>
>> Matthew Whalley wrote:
>>> Hi All,
>>> Can anyone tell me how SPM implements multiple masks? I've tried two
>>> methods which I'm told should produce equivalent results:
>>> The first method is to select the main contrast, then select three
>>> other contrasts as masks in the spm GUI (by ctrl+clicking), and
>>> exclusively masking them at p=0.05
>>> The second method has been to use imcalc to create one big exclusive
>>> mask using the formula:
>>> i1.*((i2<threshold)&(i3<threshold)&(i4<threshold))
>>> with the two-tailed t-threshold of 2.04 (df=31, so p=0.05)
>>> Then to use this big mask to exclusively mask the main contrast
>>> These two methods actually give slightly different answers - slightly
>>> different cluster locations (some identical, some similar), cluster
>>> sizes approx 20% bigger for method 1.
>>> Which method would you go with?
>>> Best wishes,
>>> Matthew
>>> --Dr Matthew Whalley
>>> Research Fellow
>>> Department of Psychology
>>> University College London
>>> WC1E 6BT
>>> Tel: 020 7679 5365
>>> Email: [log in to unmask]
>>
>
>
|