[VoxBo] vbdumpstats & vbview

[Daniel Y Kimberg]

Ling Wong wrote:
> I'm trying to use vbdumpstats on a second-tier glm. After making 2
> RFX scripts for a single cue effect (one for each timepoint), and
> then making a second-tier GLM on those tes files (to see changes
> across time), what is the difference between the "all scale" stat
> and the "first scale" stat in the vbdumpstats output?  I didn't
> specify a contrast when running vbdumpstats because it was already
> part of the glm.

Just for reference, here's the same thing from one of my test GLMs:

[./s01] contrast=all scale=t effdf=163.63 stat=967.592196301073 one-tailed_p=0.0000 two-tailed_p=0.0000
[./s01] contrast=first scale=t effdf=163.63 stat=-1.948462604522 one-tailed_p=0.9735 two-tailed_p=0.0531

When you don't specify a contrast, vbdumpstats reports the statistic
in the specified region for every contrast in your contrasts.txt file.
If you don't have a contrasts.txt file, or if you have the default
file, it reports two contrasts: "all" and "first."  This is just
shorthand for a contrast that weights every covariate of interest 1,
and a contrast that weights just the first covariate 1 (the rest 0).
These contrasts are weightings of your parameter estimates (not the
same thing as the condition contrasts you might have used in building
your G matrix).  Right now there's no way to specify this kind of
contrast as part of running the GLM, so you may want to edit the
contrasts.txt file.

> Also, for individual GLMS, should I use the -d and -n flags with vbdumpstats
> if I detrended and mean normed in the GLM, or can I exclude the flags with
> the new patch?  When did this change take place? (I didn't add the -d and -n
> in a previous analysis, when I probably should have)

The latest code should detect whether these flags were used in the
original GLM and do the same thing, so it should be safe to omit them.
But vbdumpstats will still tell you what it's doing, so you can
double-check.  I can't remember exactly when this was fixed, I believe
it was discussed on the mailing list.

> And finally, I've been having trouble loading my prm files in
> vbview.  For a while I thought it was just my GLMs for some reason,
> but the same thing happens when I use someone else's prm.  Is
> anybody else having this problem?

I haven't heard anything about this, and it seems to work for me.
Email me the path to a file that's causing problems and I'll check it
out.  I have been installing new versions of vbview at the CfN a bit
lately and it's possible I broke something.

dan