Hi,
 
I'm a new SPM user, and I'm trying to analyze single-subject data from
an event-related fMRI study with randomized trials.
 
In one analysis, I have two conditions that are randomly intermixed, and
I've built a regressor using the starting numbers of image files for the
images in each condition in the vector of onsets.  For example,
condition A might include trials starting with image numbers 5, 9, and
17.  Condition B might include trials with image numbers 1, 13, and 21.
Each trial includes 4 BOLD reps with a TR of 2.5 s, so each trial is 10
s long and covers 4 image files, starting with the one indicated in the
vector of onsets.  In all, there are about 1500 images.
 
My problem is that I'm trying to check my assumptions and see if the
hrf/basis functions that I've specified are the right ones for the
data.  When I try to plot fitted responses or anything to do, it seems,
with plotting the modeled response vs. the actual response of the voxel
over time, I get this error:
 
On line 158  ==> [Ic,xCon] = spm_conman(xX,xCon,'T|F',Inf,...
 
Error in ==> /data3/spm99b/spm_results_ui.m
On line 238  ==> [SPM,VOL,xX,xCon,xSDM] = spm_getSPM;
 
 
SPM99b: spm_results_ui (v2.22)                     17:27:10 - 17/01/2000
 
===========================================
 
 
        SPM99b: spm_getSPM (v2.20)                 17:27:10 - 17/01/2000
 
        ----------------------------------------------------------------
 
        contrast structure              :           ...saved to xCon.mat
 
        SPM computation                 :                        ...done
 
??? Error while evaluating uicontrol Callback.
 
 
>> ??? Index exceeds matrix dimensions.
 
Error in ==> /data3/spm99b/spm_graph.m
On line 391  ==>                j      =
1:size(Sess{s}.sf{t},2):length(Sess{s}.ind{t});
 
Can someone help me?
 
I also have a few other questions relating to my analysis.
 
What happens if I use a high-pass filter, when my trials are
randomized?  Am I filtering out
experimental effects that happen on slower time scales if I use a filter
of 2x my trial length?
 
If I have 2 regressors in my design matrix, what does the t-contrast 1 0
model?  This is the actual fit of that regressor to the data in terms of
explained variance, right?  Does 1 -1 test whether regressor 1 fits
better than regressor 2, then, or is it testing something else?  What if
I just want to test for greater activity during condition A than
condition B, irrespective of the fit of the regressor for each condition
to the data overall?
 
If I have several regressors, and some of them are 0's in the contrast
that I specify, is it simply leaving those regressors out of the model,
so that it would be equivalent to a model that does not include those
regressors at all, or is it somehow regressing out activity related to
the 0 columns and testing the fit of the regressors I have specified in
the contrast (with 1 or -1) with the residuals?
 
If I have 6 conditions, A B C D E F, and I want to test for voxels that
have greater activity in A and B than in C D E F, can I do this by with
contrast weights 2 2 -1 -1 -1 -1, as I would in a behavioral contrast?
 
Can I export the fitted and actual responses at a given voxel in
numerical format, so that I can test
them outside of SPM?
 
Thanks for your help!
Tor Wager
 
_____________________________
Tor Wager
Department of Psychology
University of Michigan
Cognition and Perception Area
525 East University
Ann Arbor, MI 48109-1109
 
Office: 734-936-1295
Home:   734-995-8975
Email:  [log in to unmask]
_____________________________