Hello -
I have a question about smoothness of datasets. I have a bunch of individual
subjects whose DLH is between 0.3 - 0.7. Judging from playing around with
this stuff a bit, higher numbers indicate less smoothness, which increases
the p value of any one cluster that might survive thresholding (since there
are fewer resels, I guess).
However, in my group analysis on these subjects, the DLH drops by an order of
magnitude to 0.0366. Is this normal? Is there something I can do that will
prevent this apparent increase in smoothness? Obviously, averaging a bunch
of datasets makes the result smoother, but I guess that since my subject data
had already been smoothed during preprocessing (6mm FWHM), I hadn't expected
such a dramatic change. Of course, the end result is that the cluster
statistics on my group data are very weak.
I noticed also that the computation of the smoothness estimate is dependent on
the dof. Given the nature of my design, (a single 2nd level analysis across
all conditions & subjects implementing a full factorial 3x5 ANOVA), my dof
are extremely high (240 inputs - 30 EV's = 210 DOF). I'm wondering if
perhaps a cluster based stat is misrepresentative in this case.
Alternatively, it seems to me that perhaps the true DOF for any one contrast
should be much less than this. Typically, in an ANOVA the DOF for any one
effect or contrast is only a subset of the overall DOF.
Any input would be appreciated!
p.s. please cc: me directly on any responses so I don't have to wait for the
digest
Ed
--
Ed Vessel
U. of Southern California [log in to unmask]
Dept. of Neuroscience
HNB, 3641 Watt Way http://geon.usc.edu/~vessel
Los Angeles, CA 90089-2520
(213) 740-6102
|