Dear Karl, JB and others:
I am confronted with a very similar problem. We studied a single case
afflicted by a fairly rare neurological and predominantly motor
disorder. Our hypothesis concerned the cortical sensory map, i.e. on the
basis of neurophysiological lines of evidence we expected a significant
difference in the spatial extent between the left and right hemispheric
somatosensory cluster along the postcentral gyrus. In addition, we did
expect a significant deviation of the activation centroids Talairach
coordinates. Our protocol involved some left-sided, right-sided and
bilateral sensible stimulation as well as periods of no stimulation.
Thus, I obtained t-contrasts on left vs. rest, right vs. rest, bilateral
vs. rest, bilateral vs. right and bilateral vs. left. The afflicted
hemisphere does consistently present a smaller and somewhat upwardly
displaced cluster along the postcentral region (both at corrected and
uncorrected p=B4s [0.05 & 0.001, respectively] as well as with the extend=
threshold left at 0 and 6 [the latter seems to suit our data]). More
specifically, the cluster of the nonafflicted hemisphere comprises 2 to
3 times of the voxel / resel counts than the one of the afflicted. Aside
from descriptive reporting, there is no formally acceptable way to test
for the difference in spatial extent of the left- vs. right-sided
cluster, is there? (I did not yet look at SVC. Of note, a small
cerebellar and potentially another frontal cluster may also show up but
I am not really interested in them. We may be able to sample another
patient, two individuals with a different disorder of somehow alike
semiology and two healthy controls). I assume there is no way to test
for the difference in Talairach coordinates on both hemispheres either.
Any comment will be appreciated at the time of your convenience-
TIA, andreas
Jean-Baptiste Poline wrote:
> =
> Dear Eraldo,
> =
> Not if you are trying to compare *only* two clusters (say one from
> group/condition A
> to one from group/condition B). This is simply because we have good
> approximations
> of the distribution of the size under the NULL hypothesis but not under=
> the alternative hypothesis, under which you probably are ...
> Of course, things are different if you have more than one estimation of=
> this
> cluster size under gp/cond A and B ... in which case you'd know what to=
> do ...
> =
> very best
> jb
> =
> >
> > Dear Karl and other experts in SPM statistics
> >
> > It there is a formal way (according to theory of gaussian fields) of
> > comparing whether the size (number of voxels over threshold) of two
> > clusters is signifcantly different?
> >
> > Thanks for replying
> >
> > E. Paulesu
> >
> > Prof. Eraldo Paulesu
> > Universit=E0 degli Studi di Milano - Bicocca
> > Dipartimento di Psicologia
> > Piazza dell'Ateneo Nuovo 1
> > Edificio U6
> > 20126 Milano, Italia
> > phone +39 02 6448 6748
> > fax +39 02 6448 6705
> > *************************************
> > Laboratory address
> > Centro Ciclotrone/PET
> > Scientific Institute H San Raffaele
> > via Olgettina 60
> > 20132 Milano
> > Italia
> > phone +39 02 2171 7544
> > fax +39 02 2171 7558
> >
> > Web http://web.tiscalinet.it/epaulesu/
> =
> --
> =
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~+
> | Jean-Baptiste Poline [log in to unmask] |
> | 4 Pl General Leclerc |
> | CEA/DRM/SHFJ Tel +33 (0) 1 69 86 78 52 |
> | *NEW* Tel +33 (0) 1 69 86 78 37 |
> | 91406 ORSAY FRANCE Fax +33 (0) 1 69 86 78 68 |
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~+
|