I wish to focus on set-level measures in my data. I ma particularly
interested in applying the "tests for distributed, non-focal activations"
described by Worsley et al in a paper by that name [Neuroimage 2:183-94,
1995].
I believe that the set-level analysis implemented in SPM96 is similar to,
but not the same as, the MSOS [mean sum of squares] test described by
Worsley. This appears to be particularly true when k=0. As I am using
SPECT data, where f<1, this is the prefered level of k for set-level
inferences, if I understand Friston et al. [Neuroimage 4, 223-35, 1996].
My questions are:
1. Is there any information on the behavior of the set-level test
implememnted in SPM 96ff over different smoothing kernels? in the initial
paper introducing set-level analysis [Friston et al, ibid], ROC and power
analyses were provided witout regrad to the effect of smoothing the data at
several levels. In a study in which it is expected that the signals of
interest are of different widths, as in my case, I predict this information
is important.
2. Has anyone at the FIL or elsewhere ever implemented the MSOS test in
Matlab? Was the modification to the Activation Proportion test proposes in
Worsley (above) ever integrated into a previous version of SPM, or
elsewhere? I would be extreemely grateful for even guidance on how to
implement these if no such code already exists.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|