Dear NBS's experts,

I have one study with 5 groups of 10 subjects for each one. For each subject I have one connectivity matrix. I also have three covariates. I would like to compare these connectivity matrices between the 5 groups. One suggested me to choose an ANCOVA model.

But I don't know what design should have my design matrix with ANCOVA model ? Could you explain me the theoretical design of my design matrix ?
And how should be defined my contrasts to get the effects of groups ?

Many thanks in advance !

Best regards,
Matthieu

Hi Matthieu,

The design matrix for an ANCOVA with 3 groups, 2 subjects per group and one covariate would look something like this:


1 1 0 4.53
1 1 0 3.90
1 0 1 2.10
1 0 1 9.43
1 0 0 6.98
1 0 0 8.88

The first column is the intercept. The 2nd is group 1. The 3rd is group 2. The 4th is your covariate.

The contrast is: [0 1 1 0]

The test to select is "F-test"

You should be able to extrapolate this to 5 groups.

Andrew



Originally posted by Matthieu Vanhoutte:

Thank you Andrew !

But doesn't it miss a column for the group 3 ? And so the contrast would be [0 1 1 1 0] ?
Best,
Matthieu

Hello Andrew,

Could you answer me about the design of ANCOVA : But doesn't it miss a column for the group 3 ? And so the contrast would be [0 1 1 1 0] ?

Because the ANCOVA example in NBS looks like something I don't understand...

Thank in advance for helping !

Best regards,
Matthieu

If you add a column for group 3, the design matrix will be rank deficient, which will cause problems. Try it!


Originally posted by Matthieu Vanhoutte: