Dear Joe
> This question regards the proper method of setting up a blocked
> paradigm. Specifically, can a value of 0 be used to represent the
> baseline condition? Let's say I have a '2 condition' fmri experiment,
> Task and Rest, each condition lasts for 10 images, there are a total of
> 160 images, and the experiment started with Task. When I set up the
> blocked paradigm, can I use 0 to represent the Rest condtion:
>
> conditions: 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0
> epoch_length: 10
>
> With this design matrix, I get 1 contrast, which I give a value of [1]
>
> Is this design equivalent to:
>
> conditions: 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
> epoch_length: 10
>
> Here, I can submit a contrast of [1 -1] .
Yes it is. This is because the sum of the modeled effects of 1 and 2
add up to a constant and therefore explicitly including the second
regressor doesn't really change the model.
> The outputs from both of these appear to be similar. Can the first
> design matrix (with 0 for the baseline condition), be extended to a
> multisubject experiment (let's say 3 subjects, fixed-effects model)?
> With this design matrix, is a contrast of [1 1 1] equivalent to the
> design matrix of [1 -1 1 -1 1 -1] using the second method of modelling
> the conditions? I guess the question is, when can a value of 0 be used
> to model the baseline/rest/control condition?
Yes it can. However in SPM98 this option will be removed (to allow for
more versatile model specification).
> Is there some rule that the value of the contrasts should sum to 0
(which would seem to preclude modelling the baseline as 0) ?
Contrasts usually (but not always) sum to zero to ensure estimability.
Strictly speaking the contrast of effects is the compound of
explanatory variables weighted by the contrast weights (the vector we
loosely refer to as the contrast). In your example the actual contrast
for a session-specific design and contrast weights c =[1 1 1]; would be
a single column modeling the average difference between the activation
and baseline conditions over subjects. This is a perfectly allowable
contrast.
With best wishes - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|