Exceedance Probabilities for the Dirichlet Distribution

TL;DR

This paper introduces an efficient method for calculating exceedance probabilities in the Dirichlet distribution for multiple event types, improving neuroimaging model selection processes.

Contribution

It presents a novel, efficient approach for Dirichlet EP calculation and compares it to existing sampling methods used in neuroimaging.

Findings

The new method is computationally faster than sampling approaches.

It provides accurate exceedance probabilities for multiple event types.

Application to neuroimaging demonstrates practical utility.

Abstract

We derive an efficient method to calculate exceedance probabilities (EP) for the Dirichlet distribution when the number of event types is larger than two. Also, we present an intuitive application of Dirichlet EPs and compare our method to a sampling approach which is the current practice in neuroimaging model selection.