FMRIB: EEGLAB error in function fmrib_fastr() at line 914

[nucleuscub]

Issue transferred from Bugzilla (Bug 693)

Makoto 2008-11-17 19:21:12 PST
Hi arno and rami. Here is the more complicated one.

This typically happens when you subtract more than two PCs.
The error message is as follows:

EEGLAB error in function fmrib_fastr() at line 914
Index exceeds matrix dimensions

In my case, data are 329.88 sec long.
In the line 468, 'markerl' is determined as follows:

markerl = [31, 34, 34, 34, 6]

It determines 'starts' and 'lasts' from the line 521, which determines the number of rows of 'pcamat' and finally the number of columns of 'ascore' in the line 882. For the present case,

starts lasts ascore
1 29 40801x26
33 63 40801x28
67 97 40801x28
101 131 40801x27
135 165 40801x27
169 173 40801x2

Then in the line 914,

papc=double([ascore(:,1:pcs) ones(pre_peak+max_postpeak+1,1)]);

where 'pcs' is the number of PCs (for the present case, 4). Therefore,

ascore(:,1:pcs)

can not be carried out when pcs=4 and ascore=<40801x2>.

Though my understanding of fmrib_fastr is pretty suspicious, my recommended solution is to set 'SecT' in the line 176 to the different value. When I changed it from the default value of 60 to 55, the parameters in the following part changed into the following:

markerl = [29, 31, 32, 31, 32, 18]

starts lasts ascore
1 27 40801x24
31 58 40801x24
62 90 40801x26
94 121 40801x24
125 153 40801x26
157 173 40801x14

This time it worked, because the number of columns in 'ascore' is always larger than pcs (this time it was 4).

Makoto

[reply] [−]Comment 1Arnaud Delorme 2009-07-02 15:14:51 PDT
Rami is not handling this project any more. So it is up to whoever finds the problem to fix it.

Thanks a lot,

Arno