help > RE: gPPI Error
Feb 1, 2016  04:02 PM | Donald McLaren
RE: gPPI Error
(1) This is an odd message that rarely comes up for this line. I think I've seen it a few times.

Can you verify the line 369 in timeseries_extract is: y=y-spm_FcUtil('Y0',SPM.xCon(xY.Ic),SPM.xX.xKXs,beta);

If this is not the line, can you send me the line?

If this is the line, then the issue is likely xY.Ic. Insert 4 lines above, with the following:
xY.IC
SPM.xCon(xY.Ic)
size(SPM.xX.xKXs)
size(beta)

Then re-run with the output for the subject that failed.

(2) This is the result of having NaN or Inf values in y. At line ~365 and ~373, can you insert the following line:
sum(isnan(y))

This will help to determine if the NaN came in before or after the correction with the omnibus contrast. Then try running it again.

I'm a bit concerned that the VOI size is not the same for these two subjects. How did you generate the VOIs for each subject?

Best,
Donald


Originally posted by Andrew Alan:
Hello Dr. Donald McLaren,

I am having some trouble with an error I received while computing gPPI for two amygdala VOIs. 

The right amygdala VOI worked well, using a mask from pickatlas. 

However, when I try to compute gPPI for the left amygdala I get the following errors for 2 participants. 

1)
Index exceeds matrix dimensions.

Error in timeseries_extract (line 369)
Error in PPPI (line 461)

2)
Error using svd
Input to SVD must not contain NaN or Inf.

Error in timeseries_extract (line 378)
Error in PPPI (line 461)

I checked the mask.mat files created by the first level analysis, and they look the same as those created by other participants in which gPPI was successful. 

I also opened the omnibus F -test con-images created by gPPI, and set p=1. There appears to be no missing information that would result from a faulty mask.mat file created in the first level.

When I did an ROI analysis for the amygdala VOI using the omnibus F test, I got a cluster size of 221 with Z=2.71 for the participant I got an error for (another participant I got an error for had a similar cluster size but with a Z=4.21). Then for other participants, which ran properly, the cluster size was 220 with Zs of around 3.71

Any ideas would be much appreciated! Thank you for your time!

Threaded View

TitleAuthorDate
Andrew Alan Jan 29, 2016
Donald McLaren Feb 3, 2016
Donald McLaren Feb 3, 2016
RE: gPPI Error
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 1, 2016
Donald McLaren Feb 1, 2016
Andrew Alan Feb 2, 2016
Andrew Alan Feb 2, 2016
Donald McLaren Feb 2, 2016
Andrew Alan Feb 2, 2016
Andrew Alan Feb 2, 2016
Andrew Alan Feb 1, 2016
Andrew Alan Feb 1, 2016