Dear SPMers,
I'm trying to get the principal eigenvariate time series from several brain locations in SPM8. I'm using a spherical volume of interest. The resulting time series have large edge artifacts. For example, in one case, all the time series values are between 123 and 132 except for the first and last time time points which are around 165. The same effect appears in the first row of the xY.X0 output from this calculation.
When I plot the adjusted time course from this voxel, there are no edge artifacts. This even happens if I set my VOI sphere radius to 0, which the help text from spm_regions.m says should return the adjusted voxel time-series scaled to have 2-norm of 1. I compared the result of the principal eigenvariate calculation with a sphere radius of 0 and the adjusted time series. They are very similar for most of the time series. Still, the first and last values from the principal eigenvariate are much larger than the adjusted time series and time points 2-30 and 273-297 are slightly smaller than the adjusted time series (298 volumes total).
I'll note that I'm giving the program real data with added simulated responses (preprocessed outside SPM). Perhaps there's something unique about how I altered these data that SPM doesn't like. For this data, the GLM is processed and the results are presented without program in SPM. I've tried several ways to create the simulated data & modeled it after some data that is processed correctly & can't figure out what's causing the problem. That it happens even when only one voxel is used seems a bit odd to me. Before I dig too much more into the code, I was hoping someone else might have ideas about what is happening?
Thank you for your help
Dan Handwerker
|