diut: Restricted IUT
In LTLA/giut: Generalized Intersection-Union Tests

Description Usage Arguments Details Author(s) Examples

Perform a restricted intersection-union test with two sets of independent p-values

1	diut(p1, p2)

p1, p2

a numeric vector of p-values of the same length

This function performs the restricted intersection-union test on two sets of p-values. The sampling of p-values must be independent between vectors though the configuration of true/false null hypotheses need not be. Mixtures of alternative p-value distributions are tolerated, as are correlations between tests within each set.

To calculate the restricted IUT p-value, the maximum of p1 and p2 is first obtained and defined as pm. The proportion of p-values that are greater than this maximum value is determined in each set and denoted as m1 and m2. The IUT p-value is then defined as min(pm, pm*(2-m1-m2-pm)).

This function will work best with large vectors in order to obtain precise estimates for m1 and m2. Nonetheless, the estimated p-value will be an unbiased estimate of the upper bound of the error rate with respect to imprecision in m1 and m2.

Aaron Lun

p1 <- runif(10000)
p2 <- runif(10000)
head(diut(p1, p2))

p1[1:1000] <- 1e-8
head(diut(p1, p2))

p2[5001:10000] <- rbeta(5000, 1, 20)
head(diut(p1, p2))