> what type of correlation structures can we select? Is there a list?

With -corStr you can specify an autocorrelation structure such as AR1, typically used in modeling the effect estimates from multiple basis functions such as TENT.

Other than compound symmetry, one possibility is to set a generic correlation structure (symmetrical and positive-definite), for -ranEff '~1+RT' in Example 2. Most of the time compound symmetry might be the only viable option with FMRI data without any duplicates.

> How are the glt's handled with 3dLME. Are these standard t-tests that are eventually converted
> to z-scores? Or is there more involved when constructing the contrast with 3dLME?

Previously t-statistic was provided for simple post hoc tests of pairwise comparision. To extend pairwise comparisons to generic testing with any linear combinations, the z-scores are now adopted in place of the previous t-stats.

Gang



Edited 1 time(s). Last edit at 03/12/2015 04:11PM by Gang.