[VoxBo] trial averages in vbview

[Charan Ranganath]

Oops sorry, I hit send too soon. Here it goes again:
 Dear Dan and team voxbo, 
the new version of vbview is much improved, so we have been pushing it to
the limit. Here are some things we've noticed which I wanted to alert you to:  
(1) plotting of the fitted values seems to be messed up. The values that I
get in vbview look totally different than what I've gotten in the old voxel
surfer. The voxel surfer fits look more reasonable, so I trust them. 

(2) the trial averaging produces slightly different results between VBview
and voxel surfer. To compare, we averaged the data from a masked region in
both the idl viewer and vbview. I'm attaching a file that has both time
series in it. As you can see, the averages are slightly different. The
differences are much more apparent if you set the averaging window to be 10
images (not shown). I have 2 hypotheses about the problem, which are not
mutually exclusive:
(A) When the averaging window exceeds the number of images that follow a
particular event, the idl program excludes those trials. VBView may be
including those trials--I guess this because when I set the averaging
window to 10, the idl viewer tells me it has 19 trials, whereas vbview
tells me it has "20 trials (or whatever)". If I am correct, this  leads to
the question of what the program does to account for the missing data points. 
(B) It may be the case that vbview and the idl viewer each have a slightly
different way of removing covariates of no interest from the time series
data. This could explain the slight variations even when averaging a
smaller # of time points (i.e., such that the same number of trials was
shown for both the idl viewer and vbview).

Let me know what you think. Thanks! CR 
 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: comparisonofvbviewandvoxsurfer.txt
Url: http://www.voxbo.org/pipermail/voxbo-general/attachments/20061107/515d1d56/attachment.txt