Hi Yuan Cao

If I understand correctly the warning is likely indicating that your control variables encompass all subjects, but you are not including any group membership information for the "other" 74 subjects (those who not HC nor BDF). One common way to fix this is simply to create another second-level covariate for those subjects with medication (e.g. HC for controls, BDF for medication-free patients, and BDM for medicated patients), and then enter the three groups in your analysis selecting a [-1 1 0] contrast across the three groups to indicate that you want to evaluate the difference between the first two groups (also including any other control variables and assigning a 0 to their corresponding location in the between-subjects contrast)

Hope this helps

Alfonso

Originally posted by Yuan Cao:

Hi experts,

I have 147 patients (of these 75 are medication free) and 150 hc. I ran the conn to compare the alff between these two groups (147 v.s. 150). Now, I need to do a subgroup analysis between unmedicated patients (75) and hc(150). Can I just create a BDF.txt file that labels those medication-free patients as 1 and those with medication and hc as 0. Then import this .txt to second-level covariates. Then run 2nd-level group analysis? In that way, when I run group analysis, the conn will have a warning that the subject is not 297(because we initially ran group analysis between 147bd and 150hc, but now it is 75 patients and 150hc). It can be run and can give me a result, but I am not sure if it is the right way.

or should I pick the medication-free patients and all health control and create a new conn project and re-run the conn with 75 unmedicated patients and 150 hc?

I attached a screen shot of 2nd-level group analysis