> what is the difference between doing a PPI analysis and doing 2 separate simple
> correlation analysis (1 for correct and 1 for error)

Simple correlation looks for the signal similarity between the seed and the rest of the brain. It's typically for resting state because you don't need to worry about teasing apart among the different sources of the signal. Also the correlation coefficient does not reveal any information about the directionality. If you have two conditions (correct and error), how do you separate the two conditions as well as the baseline?

In addition, PPI addresses a slightly different question: under one specific condition (or the contrast between two), do the trial-to-trial fluctuations (relative to the average response under the condition) say something about the interaction effect (between the seed region and the condition) on the target region? In contrast to correlation analysis, PPI implicitly assumes that the interaction goes from the seed region to the target region.

> If PPI is preferable, would it be wise to put correct and errors in the same model (in
> the same 3dDeconvolve) if so, should the correct and errors be in the same regressor
> or as separate regressors OR should I do PPI analysis for the correct and a separate
> PPI analysis for errors and compare the interaction terms from each using a ttest.

You can run a PPI for correct vs. errors, and/or two separate PPIs for correct vs. baseline, errors vs. baseline.

Gang



Edited 1 time(s). Last edit at 03/13/2014 05:43PM by Gang.