Hello,
I'm appealing to the group's and to Dr. Worsley's expertise in this
area. This question has been passed to me:
Hi Kris: The question pertaiins to power considerations surrounding an
fMRI study involving PTSTD patients. We are using a random-effects
analysis in comparing two groups of such patients, with respect to
activation in each of 4 regions of interest (e.g., thalamus), during
their imagining of trauma-related events. Our granting agency requires
an estimate to the number of subjects necessary to achieve a desired
power for detecting specified group-mean differences. We have
preliminary estimates for these differences, and also for variance
across subjects, within groups, regarding the activation data to be
submitted to SPM.
We wish to estimate the sample size necessary to achieve the desired
power for an individual comparison, following alpha adjustment for
multiple comparisons. For example, a Bonferroni correction for a family
of 4 comparisons, with a per-family alpha of .05, would require that
each comparison be carried out at the .0125 level; the sample-size for
the desired power, in turn, would be calculated with this value in
mind. (Subjects actually are assessed at 1, 3, 4, and 12 minths-month
intervals, so the number of experiment-wise comparisons acutally is 16,
leadiing to a Bonferroni adjustment of .05/16).
I believe, however, that Worsley's fMRI adjustment-methodology would
be more conservative. We need to know what that adjustment would be, so
we can adapt our power calculations specifically to the Worsley
procedure.
Also, given an increased conservativeness of the Worsley procedure,
available Tables for the caluculation of n-size (e.g., Jacob Cohen's)
may not be inadequate. They typically only go to a value of alpha=.01.
Does anyone know of tables that might accomodate much lower alpha
levels?
Can anyone help with regards to the mentioned Worsley correction for alpha
levels and with the power tables issue?
Thanks very much in advance for your time and consideration.
Regards,
Kris Boksman
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Kristine Boksman, B.Math, M.A. The University of Western Ontario
Ph.D.1 Clinical Psychology Department of Psychology
[log in to unmask] Office: Rm 7250a
There are two ways to live your life.
One is as though nothing is a miracle.
The other is as though everything is a miracle.
Albert Einstein (1879-1955)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|