Interval estimation in a finite mixture model: Modeling PP-values in multiple testing applications

The performance of interval estimates in a uniform-beta mixture model is evaluated using three computational strategies. Such a model has found use when modeling a distribution of PP-values from multiple testing applications. The number of PP-values and the closeness of a parameter to the boundary of its space both play a role in the precision of parameter estimates as does the “nearness” of the beta-distribution component to the uniform distribution. Three computational strategies are compared for computing interval estimates with each one having advantages and disadvantages for cases considered here.