Hi Robin
>>> If I had a five conditions fMRI experiment, and made the following two
>>> t-contrasts using spm2 :
>>> [1 1 1 1 -1] and [1 0 0 0 -1;
>>> 0 1 0 0 -1;
>>> 0 0 1 0 -1;
>>> 0 0 0 1 -1]
>>> Are they all valid? And are they the same? If not, what are they
>>> meaning separately?
>>>
>> well, they are not all valid and they are not the same ..
>> 1. When you enter a contrast, the sum of coef must be 0 and thus [1 1 1
>> 1 -1] is invalid (you can try but SPM contrast manager will refuse to do
>> that) .. if you want to oppose one condition against the others the
>> correct contrast would be [1 1 1 1 -4]
>> 2. in the second case you test if any of these contrasts (or
>> combinations) is significant .. but those ones are correct as they all
>> sum up to 0
>>
> One follow-up question. If I want to test the significancy of the first 4
> conditions only, can I define it as [1 1 1 1 0]?
>
It depends on what you want to look at .. [1 1 1 1 0] will give you the
effect of the first four condition together, which is different from the
contrast suggested before that would be here
[1 0 0 0 0;
0 1 0 0 0;
0 0 1 0 0;
0 0 0 1 0]
You can think about that as you do for any behavioral data .. the first
contrast tests if you have an effect in your expe (against the mean or
the baseline if not modeled .. see the previous message of Mauro). For
ex. with RTs it would be smthg like: is the mean over all conditions
different than my baseline (let's say pressing of button for a simple
stimulus). The second contrast tests if one of these conditions has an
effect, i.e. for RTs, is the mean of one of the condition different from
baseline. Thus I suggest to also look at each condition separately as it
is possible that only two conditions give a significant result but you
cannot know that with the above contrasts.
Hope it helps
best
- cyril
|