Hi,
When I have two groups in a vbm analysis I'd normally run cdf p2t -fitt .005 18
(18 being my degrees of freedom if I have two groups and 20 subjects). The
t-value created from this is what I then put into my randomise script (i.e.
randomise -D -i GM_mod_merg_s2.5 -o grey005_s2.5_test -d design.mat -t
design.con -n 5000 -C 2.96696 -v 5.75
where the 2.96696 is my t-value.
On this occasion however I have three groups, two of which have 16 subjects and
one of which has 10 (i.e. N=42). I've set up my design mat and con files so
that 6 t-tests are conducted. Is my degrees of freedom value then N minus the
number of groups, i.e. 42-3 or N minus the number of comparisons , i.e. 42-6?
or am I violating assumptions all over the place in doing this?
Many thanks,
Ella
Ella Mc Cabe
IMAGEN Project
Lloyd Building
Institute of Neuroscience
Trinity College
Dublin 2
+353 1 8968417
[log in to unmask]
|