[VoxBo] rfx analysis

[Junghoon Kim]

Just to confirm, this issue is also applicable to the case of using
vbstatmap, right?

Thanks, Hoon

__________________________________

Junghoon Kim, Ph.D.
Moss Rehabilitation Research Institute
1200 W. Tabor Rd.
Philadelphia, PA 19141
(215) 456-6474

>>> kimberg at mail.med.upenn.edu 06/15/06 11:17 PM >>>
marina bedny wrote:
> When creating rfx tes files vbprep does not skip over covariates of
> no interest in applying the contrast vector. So if you happened to
> add your covariates of no interest first for some subject, the order
> of the contrast vector will not be right for these subjects. Most
> people probably know this already, but I thought the info may be
> useful for someone.

Thanks for posting this.  I can barely keep track of these quirks
myself, because we've been so inconsistent about which tools require
complete contrast vectors and which just want the covariates of
interest.  In part, this is one of the reasons for introducing the
contrasts.txt file, which should work just fine with rfx scripts (use
the name of the contrast instead of a vector, no quotes needed if
there are no spaces in it).  Soon you'll be able to define your
contrasts of interest at the same time you design your GLM, and if all
goes well only ever refer to them by name (you can still edit them
later if needed).

I would definitely encourage everyone to use contrasts.txt in
conjunction with the rfx scripts, except that I haven't tried it yet.
I will try to do so tomorrow, but let me know if you have a bad
experience with named contrasts and rfx scripts.  The contrasts you
define in a contrasts.txt file should just have the covariates of
interest in them, and you should get an error otherwise.

dan