What do you want to do exactly? Test whether adding the age*group interaction is significant? If so, then you can just run model2 with a contrast looking at that term (or those terms). BTW, it would seem that the model they describe would be ill-conditioned because age*b2 and group*age*b3 would be redundant. doug
On 04/04/2013 11:04 AM, j janssen wrote:
Hi,
i read the following in a paper and would like to do the same:
"At each vertex in each hemisphere, the proportion of total variance in cortical thickness (CT) accounted for by 2 linear regression models was compared: One that did not include an age-by-group interaction term, Model 1: [CT = intercept + (group * b1) + (age * b2)] and one that did, Model 2: [CT = intercept + (group * b1) + (age * b2) + (group * age * b3)]. This resulted in an F ratio map for each hemisphere showing the degree to which the inclusion of an age-by-group term increased the proportion of CT variance that was accounted for."
i have set up the matrices for the 2 models. my questions:
- do you think that i should test the effect of age in model 1
(contrast: 0 0 1) and the effect of age * group in model 2 (contrast 0 0 0 1 -1)? if not, what do you suggest?
- do you have any idea how to combine these two analyses in a
resulting F ratio map?
thanks, -joost
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer