Dear Hengyi,
everything is fine in your model, but for comparing two samples the "One way Anova (with
constant)" option in "Basic models" is easier to use.
For modulated data it is (depending on your hypothesis) recommended to include overall GM or
GM+WM estimates as nuisance parameter in your model. Modulated data are corrected for volume
changes due to spatial normalization. This includes linear (scaling) and non-linear (warping)
components. If you are interested to compare your samples while controlling for whole brain
volume or whole GM volume you can define GM+WM or GM volume as nuisance parameter in the
Ancova model. Another option is to use the overall volumes as user-specified globals with the
"global calculation" option. The difference between both approaches is that the Ancova model
removes variance explained by overall volume from your data and the global scaling option relates
your local GM estimates to overall volume. The decision about what to use depends on your
hypothesis and I would follow the same guidelines as in conventional volumetry. You can load the
raw volumes which are saved in vbm_volumes.txt with [GM, WM, CSF] = cg_read_vbm_volumes.m
(http://dbm.neuro.uni-jena.de/cg_read_vbm_volumes.m) and GM or GM+WM can be used as
overall volume.
The non-sphericity correction for comparing groups allows to model unequal variances for each
sample (which is often true for comparing patients with controls). In this case the covariance
components are estimated for (replications over) subjects and there are no correlated repeated
measures, thus select "no" for the question about correlated repeated measures.
Best regards,
Christian
> Hi, all?
>
> I am doing a VBM study to compare a patients group and a control group
> by SPM2.
> I have used VBM scripts from Christian Gasers
> (http://dbm.neuro.uni-jena.de/vbm.html) and generated the optimized
> segemented gray matter files. If i want to compare the GMV (gray matter
> volume), i should use the mGwxxx files (modulated file), but if i want
> to compare GMC (gray matter concentration/density), i should use Gwxxx
> (non-modulated) files directly, right?
>
> And, i am using the PET model to do the statistics (Compare-populations:
> 1scan/subject, AnCova). I used the following parameters:
> two groups (24 controls vs. 25 patients),
> two nuisance variables(age, gender),
> no global normalisation,
> no grand mean scaling,
> absolute masking(threshold = 0.1),
> omit global calculation,
> yes for non-sphericity correction,
> relications over subject(25),
> yes for correlated repeated measures.
>
> Are these parameters right? What are the meaning of relications over
> subject(25) or over group (2), and the meaning of correlated repeated
> measures?
>
> Thanka a lot for any suggestions.
>
> best,??
>
>
> ----------------------------
> Hengyi Rao, Ph.D.
> [log in to unmask]
> Department of Neurology
> University of Pennsylvania
> 3 W Gates
> 3400 Spruce Street
> Philadelphia, PA, 19104
> ----------------------------
>
|