Hi,

I managed to run through all the tutorials in the manual without any error messages, and in the attention tutorial the traditional PPPI output matched the batch script output, and I also managed to get the same results running the PPI through the GUI. This made me think I had installed correctly. However, I have run the PPI on two subjects from my own study using both the GUI and PPPI traditional method (as well as gPPI) and for both subjects the two traditional models do not match each other. The PPI.P regressor is highly correlated between two versions  (e.g. r = .96), but from 4th value onwards it differs by consistently around 0.57. The PPI.Y equivalent regressors are also quite highly correlated (e.g. r =.81) between the two versions although here the differences are more variable. However, the interaction regressors are very different (only a low correlation e.g. r = .28) and hence the models give completely different results. I am really baffled and have been trying to work out all week what went wrong. I used the same VOI.mat files for both models (GUI and PPPI) and have double checked that the same scans were used. I guess it will be hard for anyone to help me with this remotely but as I am really stumped I thought I would post just in case someone had any ideas. I ultimately plan to use the gPPI, but as the gPPI also gives radically different results to the traditional model I wanted to first make sure I am able to replicate the traditional results regardless of the method I am using. I've attached the two SPM.mat files and the VOI file used in each case.

Thanks!

Sophie

I know that there has been some changes to how SPM calculates the PPI regressors, these have not been integrated with the gPPI toolbox as I have not had time to verify that I agree with those changes.
Originally posted by Sophie Forster: