Dear Rali,

The F-contrast you'd like to use uses the 1st and the 5th of your t-contrasts, right? So, the design.fts file that specifies this F-contrast would be:

/NumWaves       6
/NumContrasts   1

/Matrix
1 0 0 0 1 0

Each F-contrast (here there is just one) uses one or more of the t-contrasts, coding those that are used as 1, and those that are not used as 0. Note that in this example are exactly 6 columns, which matches the 6 rows of the t-contrast file. Each column corresponds to a row, and the 0 or 1 indicates which should be included in each of the F-tests.

Hope this helps!

All the best,

Anderson


2012/11/27 Rali Dimitrova <[log in to unmask]>
Dear all,

Just a quick question about setting a f contrast for 3 group comparison in randomise. I apologize as I have seen similar questions been asked before but none of them answered my specific question.

I set the following design.con to look at all possible two-way differences.
/Controls Relatives Patients
/Matrix
1 -1  0
-1 1  0
1  0 -1
-1 0  1
0  1 -1
0 -1  1

When I tried to set a f contrast

/Matrix
1 -1 0
0 1 -1

I received the following error message:
'ERROR: Program failed
An exception has been thrown number of columns in f-contrast matrix doesn't match number of rows in t-contrast matrix!'

When I included only three contrasts in the design.con (a priori  hypothesis C>relatives>ill) using the same fts file, randomise ran just fine. Assuming that the f contrast uses the original EVs from the design matrix rather than the set contrasts by the design.con, I am not very sure why I received this message.

Thank you in advance.

Best wishes
Rali