Stephan (Science 03) and Passamonti (Neuroimage 08) amongst others have taken this approach, to convey the PPI in terms of split data (with a 6 second delay from block transition timepoints). The figures give only an approximation, because many data points represent BOLD during its slow transition between one context/block and another. However, they are easy to grasp at a glance and it is not difficult to implement. James Rowe > Dear SPMers > > Below is a post of Paul Fletcher concerning the adequate interpretation of a > PPI. > To sort out the ambiguity that can arise from a + / - PPI, I would like to > go to the data and plot the regression slopes separately for psychological > contexts A and B. > > I was wondering how to sort my ROI values depending of psychological > contexts A and B given that the psychological variable is convolved ? > Correlating deconvolved responses would be easier but I suppose not > extremely valid. > A third solution could be to select arbitrarily the value of the nth volume > which follows the onset of an event and corresponds approximately to the > peak of the HRF, but this solution seems a bit crude. > > Any suggestion welcomed, > > Thanks a lot, > > Swann Pichon > > > > http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind05&L=SPM&P=R276113 > <http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind05&L=SPM&P=R276113> > > Hello, > > If you imagine that a PPI arises from a significant difference in the two > regression slopes (of Y on X) under conditions A and B, the positive PPI > would suggest that the slope is more positive for A than B (put simply, X > predicts Y to a greater extent under condition A than B, although this is > ambiguous - see below). The negative PPI suggests that the slope is greater > for B than A. Of course (as with a positive PPI) this might arise for a > number of reasons e.g.: > > - X predicts Y negatively under condition A and positively or less > negatively or not at all under condition B - or perhaps X predicts Y > positively under condition B and less positively or negatively or not at all > under condition A -or any combination that could produce a more positive (or > less negative) slope for regression X-Y under condition B than A. > > > Plotting the regressions slopes for A and B separately helps to clarify the > true state of affairs. Good luck > > Paul > > On Jun 9 2005, Amit Etkin wrote: > > >> A simple question... >> >> how do you interpret a negative PPI interaction? >> >> (i.e. if you have conditions A and B, and region X, the interaction of A-B >> > with activity in X shows a significant negative interaction with activity in > region Y) > >> Amit >> >>