On Sat, Jan 7, 2012 at 12:47 PM, Jonathan Smallwood
<[log in to unmask]> wrote:
Hi SPM Masters
I have a categorical within participants factors with two levels and several continuous variables. I would like to be able to assess (i) the main effect of the continuous variables,
Repeated-measures ANCOVAs should not be used. The field of statistics hasn't come to a conclusion on the best approach. Thus, if you want to examine the continuous variances, then you need to do it separately for each level of your task with two models.
(ii) their interaction with the contrast
There is no such thing as the interaction of a contrast and a variable. I suspect that you are asking is there an interaction of the difference in level1 and level2 and the continuous variable. If so, create a difference image of level1-level2 for each subject and use simple linear regression.
and (iii) their interactions with each other.
Zero mean each variable, multiple them together and then use the multipled variable in the model in addition to the original variables. The analysis will be the same as (i) and could even replace the model in (i).
in spm 9 using the regression interface in the GUI what would the contrasts look like to do this?
You need at least 3 different models to answer these questions, but with the simple models, then it should be clear as to what the contrasts are. Form your null hypothesis (e.g. A=0). If the equation does not equal 0, make it equal 0 (e.g. A=B becomes A-B=0). Now use the 1s and -1s to form you contrast based on the null hypothesis.
Thanks
Jonny
