|
Abstract:
|
Some recent studies have compared subjects from different age or diagnostic groups on measures of resting-state functional connectivity. The responses being compared across groups are large correlation matrices, representing connectivity among a set of brain regions. Given these non-standard responses, the null hypothesis of no group difference must be tested by non-standard methods. We shall discuss two permutation testing approaches. The first derives a test statistic from the distances among observations, an approach that has received two disparate formulations which we show to be equivalent in key special cases. The second is a graphical testing approach incorporating feature selection, which may be extended to a much broader class of high-dimensional testing problems.
|
