> From [log in to unmask] Tue Nov 27 18:47:39 2001
> X-MSMail-Priority: Normal
> X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700
> Date: Tue, 27 Nov 2001 10:45:23 -0800
Dear Yunyu,
> I asked this question on the spm mailing list last week. But I haven't got
> any answer yet. Maybe somebody could help with this question.
I am not sure if anyone has answered yet - if they have not:
> I specified a model for my fmri data set in SPM by using 3 conditions A, B
> and C, besides there is a neutral condition (which was not specifically
> modelled). So the paradigm is: neutral(45 scans) -- A(60 scans) --
> neutral(48 scans) -- B(56 scans) -- neutral(51 scans) -- C(55 scans) --
> neutral(50scans). The TR is 6 seconds and since each condition appeared only
> once, the SOA for each condition was specified as 353 scans (or any large
> number >=353 scans should do it).
> Then, I selected "no" for stochastic design, "variable" for SOA (but it
> really doesn't matter even if the SOA is fixed), input the
> time scans and durations; took "epochs" and "box-car" for the design; set
> hrf "no". The idea was not to use any hemodynamic response function, since
> I wanted to define a pure box-car function for each condition. When the
> design matrix was plotted, the functions for A, B, and C
> had a triangular appearance instead of a boxcar appearance. With triangular
> appearance, I mean that the condition for A as a function of time was
> linearly increasing up to a maximum and then decreasing to zero for the
> stimulus duration. This behavior was not expected, since a boxcar i.e.
> square wave-like shape was desired for the stimulus duration. For the other
> regressors (B,C) similar shapes were obtained. As an example, you will find
> a graph in .gif format attached.
> So, my question is why are the regessors not boxcar-like since no hrf
> function was used?
I think the problem is that you selected "variable". This invokes an
event-related model of epoch-related responses with a train lasting for
the specified [variable] duration (this train has accumulated to give
the triangular regressor). Simply ask for "fixed" and everything
should be alright.
I hope this helps - Karl
|