Dear SPM users,
I have a question concerning the setting of my
height threshold value. In a first experiment (using a random effects
model),
with 10 male and 10 female subjects, height threshold was set
at p<0.0005, uncorrected, for a priori ROIs. A (non
surprising) gender difference was detected in
a specific region. A second study was conducted specifically to
confirm this gender differential
activation with 14 NEW participants
(7 male and 7 female subjects). This time however, given the above, the
ROI of interest was
limited to a relatively small region...and acceptable
height threshold was calculated as a function of the limited size of
this region.
To target a corrected p<0.05 level, an uncorrected
p<0.019 level was therefore calculated as adequate.
However, for some reason (too long to explain here), it would
be nice to combine subjects from both studies and to reanalyze this
data. How should I set my new height threshold
if I analyze the data for all 34 subjects? I know it should be
somewhere
between p<0.0005 and p<0.019 but the precise value
is hard to calculate. I was considering using a simple linear
interpolation between p values, taking into account number of
subjects in each study, to provide an APPROXIMATE value for the
height threshold. Any comments?
Sherif