This package implements the Renyi Outlier Test arXiv:2411.13542, designed for modern large scale testing applications, especially those where prior information available.
The test combines a vector of independent uniform p-values into one p-value with power against alternatives where a small number of p-values are non-null.
The test can leverage prior probabilities/weights specifying which variables are likely to be outliers and prior estimates of effect size. The procedure is fast even when the number of initial p-values is large (e.g. in the millions) and numerically stable even for very small p-values (e.g. 10^-300).
For documentation and installation, please visit ryanchrist.r-universe.dev/renyi.