Dear Fred,

Typically you would/should have entered 2 as the number of sessions per subject (and 16 as the number of subjects), and then define those "pre" and "post" acquisitions as two separate conditions (in Setup.Conditions), where you associate the "pre" condition with Session1 and the "post" condition with Session2 data (see for example these posts for additional info https://www.nitrc.org/forum/forum.php?th... and/or https://www.nitrc.org/forum/forum.php?th...). That would have allowed you to run this paired t-tests in the second-level anaysis tab simply by selecting 'AllSubjects' (in the 'subject effects' list) and selecting both the 'pre' and 'post' conditions (in the 'conditions' list) and enter a between-conditions contrast [-1 1].

Since you have entered those two conditions as if they were different subjects (32 subjects with one single session each, instead of 16 subjects with two sessions each) you can still run those paired t-tests but they are a bit more cumbersome to define.

Step 1) You need to first create a series of second-level covariates (one for each of your 16 "real" subjects), dummy coding which are the pre- and post- scans associated with each subject. For example, if you have entered into CONN first the pre data for the 16 subjects and then the post data for the 16 same subjects, then you could define a new second level covariate (e.g. named "SubjectEffects"), and enter in the 'values' field:

  kron([1 1], eye(16))

(and then answer yes to the question regarding expanding this covariate into multiple ones). This will create 16 new covariates (named SubjectEffects01 to SubjectEffects16) each pointing to the appropriate scans for each "real" subject. Note that if, instead, you have entered into CONN first the pre and post data for the first subject, then the pre and post data for the second subject, etc., then you would enter instead in the 'values' field the following:

   kron(eye(16), [1 1])

in order to create the appropriate SubjectEffects* covariates.

Step 2) After these you would define a 'pre' second-level covariate dummy coding all of the pre scans (e.g.kron([1 0],ones(1,16)) in the first case above, or kron(ones(1,16),[1 0]) in the second case above) and another 'post' second-level covariate dummy-condign all of the post scans (e.g. kron([0 1],ones(1,16)) in the first case, or kron(ones(1,16),[0 1]) in the second case)

Step 3) Last, in the second-level analysis tab, you would select 'Pre', 'Post', and the 16 'SubjectEffects*' and enter the contrast [-1 1 zeros(1,16)] in order to perform the desired paired t-test (see for example this post for additional/related info for more complex designs https://www.nitrc.org/forum/forum.php?th...).

Hope this helps
Alfonso

EDIT: corrected a few typos in the second-level covariate definitions

Originally posted by fred Sampedro: