Hi Anderson, 

I noticed my designmatrix shifted, so attached you’ll find another try:

Hi Anderson,

Thank you! I am, however, not sure if I understand you correctly: do I run two separate higher analyses: one as you suggested to test for the mean effect of my first level contrast, and a separate paired samples t-test or do I combine your suggestion with my design into something like this (see attached)

Best,

Sandra

Op 5 dec. 2016, om 08:35 heeft Sandra Thijssen <[log in to unmask]> het volgende geschreven:

A. 

1 1 1 0 0

1 1 0 1 0

1 1 0 0 1

1 0 -1 0 0

1 0 0 -1 0

1 0 0 0 -1

mean effect C1: 1 0 0 0 0

H1 > H2 C2:  0 1 0 0 0

H1 < H2 C3: 0 -1 0 0 0

 

Or B:

1 1 1 0 0

1 1 0 1 0

1 1 0 0 1

1 -1 1 0 0

1 -1 0 1 0

1 -1 0 0 1

mean effect C1: 1 0 0 0 0

H1 > H2 C2: 0 1 0 0 0

H1 < H2 C3: 0 -1 0 0 0

Thanks again!

Sandra

Op 3 dec. 2016, om 10:49 heeft Anderson M. Winkler <[log in to unmask]> het volgende geschreven:

Hi Sandra,

Yes, it's possible, although the design looks different for C1. Something as this:

1       1       0       0

1       0       1       0

1       0       0       1

1      -1       0       0

1       0      -1       0

1       0       0      -1

And the contrast then is:

C1: 1 0 0 0

Your current design is fine for the paired t-test (C2 and C3 in your list), but ideally you'd remove EV2, as that makes the design rank deficient (in reality it won't affect in this case, although this depends on the software implementation than on the maths).

All the best,

Anderson

On 2 December 2016 at 08:51, Thijssen, S. <[log in to unmask]> wrote:

Dear experts,

 

At our lab, we’re running a within-subject hormonal administration study. In this study, participants perform the same fMRI paradigm twice, once while receiving hormone H1 and once while receiving hormone H2. We’re trying to set up our higher level analysis, but came up with some questions. If we perform a two-sample paired t-test (as described on https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FEAT/UserGuide#Paired_Two-Group_Difference_.28Two-Sample_Paired_T-Test.29), we can test for the differences between H1 and H2 on our first level contrast. However, we are also interested in a mean group effect of our first level contrast, irrespective of which hormone was administered. We were wondering if it is possible to combine designs for unpaired and paired two-sample t-test in order to assess both the mean effect and between session differences. Below I have written out our design matrix. Does this look OK or is there a different way to test both the mean effect and between session differences of a first level contrast?

 

Thank you very much for help!

 

Best,

Sandra

 

EV1-H1  EV2-H2  EV-PP1  EV-PP2 EV-PP3

1       0       1       0       0

1       0       0       1       0

1       0       0       0       1

0       1       1       0       0

0       1       0       1       0

0       1       0       0       1

 

C1      0.5     0.5     0       0       0

C2      1       -1      0       0       0

C3       -1      1       0       0       0