Excellent, thanks a lot! In paired t-test I had not understood originally that I need to enter pairs of images each time. The results look indeed identical (see attached) when I use in second-level paired t-test to compare A > baseline vs. B > baseline or the contrasts A > B using one-sample t-test. My baseline is the same in both cases (my experiment includes A, B and baseline condition). On Wed, Aug 17, 2016 at 11:45 PM, Angstadt, Mike <[log in to unmask]> wrote: > John, > > What was wrong with the paired T-test model? In this case, that would be > the appropriate model to use given your pre-existing contrasts, and it > should tell you where A and B are different. The model itself will look odd > if you haven’t viewed design matrices like that before, as you’ll have 2 > columns for you contrasts and a bunch of columns for your subjects. > > Generally, if you have access to the original data, it would probably make > more sense to just directly contrast A and B at the first level and enter > that contrast into a one-sample T-test. But ultimately it will be an > equivalent model (assuming that baseline_condition in your contrasts is > indeed an identical baseline for both the A and B contrasts) because a > one-sample T-test on a difference of conditions is identical to a paired > T-test on the individual conditions. > > -Mike > > -- > Mike Angstadt > Research Computer Specialist / PANLab Lab Manager > Department of Psychiatry / University of Michigan > (734) 936-8229 > > > From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On > Behalf Of John Gelburg > Sent: Wednesday, August 17, 2016 11:47 AM > To: [log in to unmask] > Subject: [SPM] simple question on random-effects second-level analysis > > Hi all, > > I have two sets of con* images from two first level contrasts: condition_A > > baseline_condition and condition_B > baseline_condition. At the second > level I need a contrast condition_A > condition_B. Is it achievable given > data that I have? If so, which model should I define? - I tried a paired > t-test but it seems to make something different. > > Many thanks, > John > > ********************************************************** > Electronic Mail is not secure, may not be read every day, and should not > be used for urgent or sensitive issues >