Hi NBS users',
I would like to know how to make a two-way repeated measures ANOVA (disease, time) in my design matrix. It's not describe on Fsl website.

Information about the design:
18 subject (9 healthy control and 9 patients)
Longitudinal study (time A and time B): repeated level.

Thank you in advance for any kind of advice.
Kind regards,
Frederic

Hi Fredric,

Here is a link to some nice slides on modelling a 2x2 repeated measures (slides 78-79):

http://mumford.bol.ucla.edu/setting_up_d...

Here are some detailed instructions:

Suppose the data is order according to the following rows (2 groups, 2 time points, 2 subjects per group):

Group 1 Time 1 Subject 1
Group 1 Time 1 Subject 2
Group 1 Time 2 Subject 1
Group 1 Time 2 Subject 2
Group 2 Time 1 Subject 3
Group 2 Time 1 Subject 4
Group 2 Time 2 Subject 3
Group 2 Time 2 Subject 4

The your design matrix would look like:

1 0 0 0  1 1
0 1 0 0  1 1
1 0 0 0 -1 -1
0 1 0 0 -1 -1
0 0 1 0  1 -1
0 0 0 1  1 -1
0 0 1 0  -1 1
0 0 0 1  -1 1

Where the first four columns model the individual means of each subject, the 5th column models the main effect of time and the 6th column models the group x time interaction.

The contrast for the main effect of time is either: [0 0 0 0 1 0] or [0 0 0 0 -1 0]

The contrast for the interaction is: [0 0 0 0 0 1] or [0 0 0 0 0 -1]

The exchange block should be set as: [1 2 1 2 3 4 3 4]

Select either t-test of F-test, as appropriate.

If you want to test the main effect of group, then a repeated measures design is not particularly useful and can be eliminated. For example you can average across the two time points per subject and perform a two sample t-test, or perform a one-sample test on the difference.

Andrew

Originally posted by Frederic Briend:

Thanks for you reactivity Andrew.
Your advice and link was very fruitful.

One additional question, what is a minimum number of node advised to use your NBS software?

Best,

Frederic

Hi Fredric,

There is no theoretical minimum number of nodes (well, at least 2). If you have less than 5 or so nodes, you are probably better served by using the FDR or strong control of the FWE. But 5 or so nodes is not really sufficient to comprehensively characterize whole-brain networks anyway.

Anrew


Originally posted by Frederic Briend: