hi,
this question regards setting up a higher level model which i am basing on
the exaple of a triple t-test. except, my case has 15 conditions instead of
3. i have 8 subjects, but lets just look at 2 for simplicity sake (all in
same group). I organize my EV's as follows:
EV1 EV2 EV3 EV4 ... EV15 EV16 EV17
S1C1 1 1 1 1 1 1 0
S2C1 1 1 1 1 1 0 1
S1C2 -1 0 0 0 0 1 0
S2C2 -1 0 0 0 0 0 1
S1C3 0 -1 0 0 0 1 0
S2C3 0 -1 0 0 0 0 1
...
S1C15 0 0 0 0 -1 1 0
S2C15 0 0 0 0 -1 0 1
now, if i want to know the mean effect across all subjects for condition 1,
i will use the following contrast:
mean(C1) = [1 1 1 1 .... 1 0 0]
for condition 2:
mean(C2) = [-1 0 0 0 .... 0 0 0]
for condition 3:
mean(C3) = [0 -1 0 0 .... 0 0 0]
and so on. Now,if i want to test if the sum of the first 5 conditions is
greater than the sum of the last 5 conditions, i will use:
sum(C1:C5)-sum(C10:C15) = [0 0 0 0 0 1 1 1 1 1 2 2 2 2 2 0 0]
so, my question is firstly, is the above correct? and secondly, how do i set
up a contrast to test if there is a linear increase (or decrease) as go from
condition 1 to condition 15? i'm confused regarding if this should be
demeaned in any way?
i.e. 7*C1 + 6*C2 + 5*C3 + 4*C4 + 3*C4 + 2*C5 + C6 - C8 - 2*C9 - 3*C10 -
4*C11 - 5*C12 - 6*C13 - 7*C14 - 8*C15
to form the correct contrast, or something similar? perhaps there are other
good options....
thanks,
vitaly
|