Hi Aaron,

The contrast depends on how the design matrix was constructed. If you follow the example in the page, there would be 1 regressor per subject, plus 3 more regressors encoding the differences between the 4 conditions. With 13 subjects, the design matrix will then have 16 EVs, and the contrast as shown will have 3 rows and 16 columns.

The 4 level is omitted because this design looks for differences between the 4 levels. Once 3 differences are known involving all 4 levels, then all possible pairwise differences are automatically known, hence don't need be represented.

There are other ways in which the exact same model can be written, including one in which there is one EV per condition (so 4 overall), but then one less EV for the participants (so 12 in your case, instead of 13). The contrasts would look very different.

Re: missing subjects, yes, sorry, I meant missing conditions.

All the best,

Anderson

On Wed, 25 Jul 2018 at 22:11, Tanenbaum, Aaron <[log in to unmask]> wrote:

Thank you for replying. In regards to the rank-deficient matrix. This matrix I got straight from the URL I mentioned. It was meant for a 1-factor 4-levels repeated measures anova. Which is nice because that is what I am dealing with except I have 10 subjects not 2. So according to the tutorial my contrast matrix needs to be 13x3.

In this same example, is the input image a set of images that follow the same order as the design matrix and simply the mean of the subjects 4 levels minus one of the levels. It is not clear how the forth level is treated in the design matrix since it is represented by all zeros. I would assume that the forth level would be given it's own column like the other three.

Finally, I do not have missing subjects just that not all subjects have data for all four of experiment conditions. This is still good for randomise?

Aaron Tanenbaum

From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Anderson M. Winkler <[log in to unmask]>
Sent: Wednesday, July 25, 2018 8:08:41 PM
To: [log in to unmask]
Subject: Re: [FSL] Within-Subject Anova

Hi Aaron,

The contrast in the randomise manual is the one you should use: it will capture both positive and negative effects as the F-test is bidirectional (two-tailed).

The alternative you suggest won't work I'm afraid as that contrast is rank-deficient, i.e., it has rank 3, not 5 or 6, meaning that some of the lines don't test anything extra that isn't already present in the others (well, it may in fact work in randomise because behind the scenes it removes redundancies but this is because the implementation is nice with the user; the same may not work in other software).

Regarding missing subjects, it also works in randomise as in the example, just make sure each subject is in their own exchangeability block.

All the best,

Anderson

On Wed, 25 Jul 2018 at 14:43, Tanenbaum, Aaron <[log in to unmask]> wrote:

I have two questions. First, I am trying to do a repeated measures anova. I have a few subjects that have missing time points. I was wondering can randomise handle this. Second, in terms of the contrast file using https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise/UserGuide as an example. In this webpage it says the contrast file for the repeated measures anova should be

0 0 1 0 0
0 0 0 1 0
0 0 0 0 1

However, this only captures positive differences from the reference level. Because, I am looking for both positive and negative difference i was thinking the contrast file should look like.

0 0 1 0 0
0 0 0 1 0
0 0 0 0 1

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

Am I getting this correct or am I completely wrong?

Aaron Tanenbaum

The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.