[VoxBo] threshold calculator in IDL vv

[Joonkoo Park]

Great! Thanks a lot Dan.
I hadn't notice the command vbtcalc before. Thanks!

Joon


On 9/19/07, Daniel Y Kimberg <kimberg at mail.med.upenn.edu> wrote:
>
> Joonkoo Park wrote:
> > I want to clarify a couple of things about the threshold calculator
> > in IDL voxbo view. I often use the old vbview (from IDL) command
> > because it is sometimes more informative. The threshold calculator
> > window has data entry fields #vox, FWHM, k, eff df, and alpha. I
> > believe that all these parameters are used to calculate a critical t
> > value.
>
> As an aside, I don't think k is used.  I'm not 100% sure, but I
> believe it's part of an RFT calculation that was later decided to be
> too flaky to implement.
>
> > As far as I understand, eff df is the temporal effective degrees of
> > freedom (is this correct?). If this is the case, then why is eff df
> > the same for all voxels when each voxel may have totally different
> > time series structure of BOLD signals? I am probably confused
> > because I don't know the underlying statistics behind it, but at
> > least what I'd like to confirm is that eff df is a single value
> > across all voxels.
>
> effdf is indeed the effective degrees of freedom of the time series,
> and it's calculated in such a way that it's the same for each voxel.
> You don't really want to let your experimental effects contaminate
> your assessment of the independence of your observations, and in any
> case you wouldn't want to estimate the number of independence
> observations you have from that same set of observations.
> Fortunately, we can do a pretty good job characterizing the
> non-independence.  In VoxBo, the two things that contribute to this
> are the 1/f model and any temporal filtering you do.  Neither of these
> differs across the brain.
>
> > Another thing. My critical t value in the threshold calculator has the
> same
> > value no matter what number of voxels I put in the #vox data field (I
> did
> > press updata button). I assumed that it would deal with multiple
> comparison
> > issues here, but critical t value stays the same whether I'm looking at
> 1
> > voxel or 1000 voxels. Is this a bug? Then, the following question is
> what
> > method does it use to cope with the multiple comparison problem?
> >
> > To illustrate, #vox=16150 FWHM=0 k=1 effdf=16.69 alpha=0.05 give me
> critical
> > t=6.851, which seems like a Bonferroni corrected t-value. However,
> > #vox=16150 FWHM=0 k=1 effdf=16.69 alpha=0.05 also give me the exact same
> > critical t=6.851.
>
> Assuming the voxel count was supposed to be different for one of
> those...  When you have FWHM=0, you will always get the Bonferroni
> value, and it should definitely be sensitive to the number of voxels.
> I would say it's possible the threshold calculator just got in a weird
> state and wasn't updating.  It's also good to know that IDL doesn't
> know you've updated a value unless you hit return after typing.  You
> might consider using the non-IDL threshold calculator, vbtcalc.
> Although it's a pain because you have to type everything in yourself,
> it's nice because it gives you both the Bonferroni and the RFT
> thresholds, no matter which is lower.
>
> dan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.voxbo.org/pipermail/voxbo-general/attachments/20070919/154942d6/attachment.html