Hi Meng, in the case of a 2x2 interaction, the F test has only 1 dof in the numerator. This makes it the same as an unsigned t-test. The interaction in intercept is just the interaction regressing out gender and age (the first row of your matrix). The interaction in slope is the 2x2xage interaction regressing out the effect of gender (the 2nd row). doug

On 08/06/2012 10:28 PM, Meng Li wrote:

Hi professor, I know that if performing the Analysis of Variance (eg. 2*2 interaction analysis), we should use the F-test, so I wonder the reason why the contrast matrix of 4G0V (http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf4G0V) use the t-test. And if I want to know whether there is an interaction between diagnosis and genotype regressing out the effect of gender and age (in either intercept or slope ??) as previous emial mentioned, that is using the F-test to detect the interaction, so whether I should use the following matrix: 0.25 0.25 -0.25 -0.25 -0.25 -0.25 0.25 0.25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.25 0.25 -0.25 -0.25 -0.25 -0.25 0.25 0.25 Thanks, Best wishes, Meng Date: Mon, 06 Aug 2012 10:01:22 -0400 From: Douglas N Greve greve@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] contrast matrix To: freesurfer@nmr.mgh.harvard.edu Hi Meng, the first row is what you want. The second row tests for an interaction between diagnosis, genotype, and age. doug On 08/04/2012 10:23 PM, Meng Li wrote:

...

Hi Freesurfers,

My experimental design includes 3 discrete factors and 1 continuous variables:

Discrete factors: diagnosis with two levels (A, B); genotype with two levels (G, T); gender (F, M)

Continuous variable: age

So I can get 8 classes: AGF, AGM, ATF,ATM, BGF, BGM, BTF, BTM. Then I want to perform 2 (diagnosis)* 2(genotype) interaction analysis with regressing out the effect of gender and age. Is the following contrast matrix correct?

0.25 0.25 -0.25 -0.25 -0.25 -0.25 0.25 0.25 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 00.25 0.25 -0.25 -0.25 -0.25 -0.25 0.25 0.25

Any help will be very appreciated.

Best wishes,

Meng