Hi Yong,

If the "controls" group was scanned twice at the same times that the "patients" group and you have a 2x2 anova design, then you would typically enter the "pre" and "post" scans as two conditions (in Setup.Conditions, associating each condition with the corresponding session/run for each subject), as well as define the two subject groups (e.g. "Patients" and "Controls") as two simple dummy-coded second-level covariates in Setup.Covariates.Second-level. Then for your second-level analyses if you want to look at the group by scan interaction (comparing the post- and pre- medication connectivity differences between patients and controls) you would do that by selecting the two subject groups (Controls and Patients) in the "subject-effects" list and entering a [-1 1] contrast there, and then selecting the two scans (Pre and Post) in the "conditions" list and entering a [-1 1] contrast there as well. Other contrast of interest would possibly be:

Main medication effects (comparing post- and pre-medication connectivity measures):
   Select Controls and Patients groups and enter contrast [1 1]/2
   Select Pre and Post conditions and enter contrast [-1 1]

Simple main medication effects (e.g. comparing the post- and pre- medication connectivity measures for your patients groups only)
   Select Patients and enter contrast 1
   Select Pre and Post conditions and enter contrast [-1 1]

OR conjunction of simple main medication effects (difference between post- and pre-medication connectivity measures in either of the patients or controls groups):
   Select Controls and Patients and enter contrast [1 0; 0 1]
   Select Pre and Post conditions and enter contrast [-1 1]

Main group effects (comparing patients and controls connectivity measures):
   Select Controls and Patients and enter contrast [-1 1]
   Select Pre and Post conditions and enter contrast [1 1]/2

Simple main group effects (e.g. comparing the patients and control groups connectivity measures after medication)
   Select Controls and Patients and enter contrast [-1 1]
   Select Post condition and enter contrast 1

OR conjunction of simple main group effects (difference between patients and controls during either pre- or post- medication):
   Select Control and Patients and enter contrast [-1 1]
   Select Pre and Post conditions and enter contrast [1 0; 0 1]

note that all test above correspond to a mixed within- between-subject 2x2 anova design, with one between-subjects factor (groups) and one within-subjects factor (medication).

On the other hand, if the "controls" group was scanned just once then you cannot truly look at the above interaction (since the "control" reference condition will be the same for both the pre- and post- medication scans; in practice this means that the (Post-medication - controls) - (Pre-medication - controls)" contrast will be just the same as the "Post-medication - Pre-medication" contrast), but you can still look at the within-subject differences in connectivity pre- and post- medication, as well as comparing each of these conditions to your control group. The way to setup this experiment in CONN is perhaps slightly more cumbersome than the above scenario. If you have 10 patients (scanned twice) and 10 controls (scanned once), you would enter the 30 scans as if they were from 30 individual subjects. You would then define three second-level covariates in Setup.Conditions defining these three subject groups (e.g. "Patients-Pre" "Patients-Post" and "Controls", dummy coded using 1/0 values identifying the corresponding scans for each group), and you would also add 10 second-level covariates identifying your individual "repeated-measures" patients (e.g. "Subject1" to "Subject10" variables, dummy coded using 1/0 values to identify the two scans associated with each subject; note that you can have CONN automatically create these 10 variables for you if you create a single covariate named "Subject" and then enter in the values field the following "[eye(10) eye(10) zeros(10)]" without the quotes). Then for your second-level analyses you can do any of the following:

Simple main medication effects (e.g. comparing the post- and pre- medication connectivity measures for your patients groups only)
   Select "Patients-Pre" "Patients-Post" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 1   0 0 0 0 0 0 0 0 0 0]

Simple main group effects (e.g. comparing the patients and control groups connectivity measures after medication)
   Select "Patients-Post" and "Controls" and enter contrast [1 -1]

Simple main group effects (e.g. comparing the patients and control groups connectivity measures before medication)
   Select "Patients-Pre" and "Controls" and enter contrast [1 -1]

OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
   Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1    0 0 0 0 0 0 0 0 0 0;  0 -1 1   0 0 0 0 0 0 0 0 0 0]

Note that the first test above (simple main medication effects) will be effectively equivalent to a paired t-test comparing pre- and post-medication connectivity measures, the next two tests will each be effectively equivalent to a two-sample t-test comparing patients vs. controls connectivity measures, and the last test will be equivalent to a manova of the between-group differences pre- and post- medication.

Hope this helps

Alfonso

EDIT: the lines above that read:

 OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
 Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 0 0 0 0 0 0 0 0 0 0; 0 -1 1 0 0 0 0 0 0 0 0 0 0]

should read instead:

 OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
 Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 -ones(1,10)/10; 0 -1 1 -ones(1,10)/10]

The reason is that, for these comparisons between patients and controls, your subject effects do contain some of the information of interest (the average response across both pre- and post- conditions for your patients) which you need for the controls-patients comparisons

Originally posted by Yong Li: