Hello there,
Maybe someone out there in SPM land can help us with our current fMRI
analyses quandry. Here is our problem: we have scanned a group of 12 young
normal controls on three drug conditions using a WITHIN-SUBJECT design
(placebo vs low drug vs high drug; same subjects for all conditions,
different days). At this point, we are analyzing the data with a
random-effect model but the model is way to severe to let out any good
stuff (there are more things out there when we look at the BOLD signals,
but it doesn't reach the significance levels we want so far).
My main point here is that I am not certain at all that we have to use
a
full random effect model since this is NOT a between-subject design such as
comparing young to aged populations, but rather a within-subject design.
Usually, in stats, the within-subject repeated design is stronger than
the between-subjects design because it allows more degrees of freedom.
However, if we use a full random effect model (and not fixed), I have the
feeling that we are treating the data as if they were from different groups
of
subjects,rendering the model too severe. I am aware that I couldn't use a
full fixed effect model, but what do the SPM experts out there think about
this problem? How can I deal with the supposed advantages of
within-subjects designs in this case when we treat data as if they were
from between?
Any and all help is greatly appreciated. Thanks in advance.
Walter S. Marcantoni, Ph.D.
Brain Imaging Group
Douglas Research Centre
Montreal, Quebec
Canada
|