my first thought is that the grey matter threshold was not equivalent for
the two analyses; with roughly equivalent smoothness estimates, I don't see
another explanantion for such divergent REsel counts off the bat.
Not clear why the height thresh T value is not consistent-- it appears you
are reporting the defualt values for the rest.
if you intend to compare these analyses, I would suggest you rethreshold
one or both to have the same extent thresh [esp since the FWHM are so
close] either by choosing one value or picking a cluster size at or above
the [nominal] resolution of your system--preferbaly phantom data, or point
source in water.
At 12:07 PM 11/28/00 -0500, you wrote:
>Dear SPM ers:
>
>When reporting SPM99 results for PET scan
>obviously T or Z of all regions above significant threshold with
>x,y,z coordinates (MNI transformed to T & T etc ... should be
>clearly stated).
>
>I obtained also for a given height threshold of p of 0.001 and a size
>threshold of k=62 for two different experiments ( first in a large
>group 42 and second in a smaller SUBSET of 15 subjects of the large
>group) the following
>
>1) T and p values for size
>n=42, df=37 , height T =3.33 and p value for k=0.038
>n=15, df=12, height T=3.93 and p value for k=0.023
>I understand (at least I think I understand) why there are
>differences (EC of different T fields with different d.f.)
>
>2) smoothness FWHM
>n=42: 11.5, 13.2, 13.8
>n=15: 11.5, 13.0, 14.8
>I understand why there are differences the way smoothness is computed
>make estimate different (partial derivatives for each variable x,y,z
>but on different sample size)
>
>3) Search volume
>n=42 S= 536352
>n=15 866784 (subset of the n=42 same threshold etc)
>It is more difficult to understand the large differences (is this a
>problem with the thresholding of my data) or a different correction
>(weighting) applied by SPM99 to the n=42 versus n=15. Basically is
>it a physical reality or a construct and how is it done.
>
>4) Resels
>n=42 R=205
>n=15 R=334
>
>Should we report all these results when we compare groups (42 vs 15) ?
>(since results between different research labs can be exchanged
>(reproduced) if all variables are known)
>
>How does the search volume and the computation of the smoothness(
>FWHM in xyz) differ between SPM96 and SPM99 since I found differences
>between both although statistical results are almost identical.
>
>Thank you for your help
>
>Didine
>
>
>
>
>--
>Badreddine Bencherif, MD
>Department of Radiology
>Division of Nuclear Medicine
>Johns Hopkins University School of Medicine
>601 N. Caroline St. / JHOC 4230
>Baltimore, MD 21287-0855
>
>Phone: (410) 614-2787
>Pager: (410) 283-2050
>Fax: (410) 614-1977
>email: [log in to unmask]
>
|