PGTukey: Multiple testing adjusted p-value (familywise error rate) in...

In baolinwu/IPAM: Inference problems in various ANOVA models

Description

Given an ANOVA model with r levels, we are interested in testing the pairwise mean differences (1 vs 2,...; 2 vs 3,... etc). We consider a weighted multiple hypothesis testing problem, where we allow different tests to get unequal type I errors, and in the meantime, we want to control the overall type I errors (familywise error rate; FWER) at a pre-specified level. For a given comparison, we compute the multiple testing adjusted p-value, which is the smallest FWER that will lead to a significant test for the given comparison.

Usage

PGTukey(nr, Qx = 2, ik = 1, W = NULL)

Arguments

nr	the sample sizes for each factor level.
Qx	observed t-statistic for a comparison
ik	index of the comparison that we want to compute the multiple testing adjusted p-value
W	the relative type I errors assigned to each comparison.

Value

alpha

multiple testing adjusted p-value (FWER)

FPi

individual test type I errors

References

Ling,S., Johnson,C., Ellermann,J., Eberly,L. and Wu,B. (2019) A generalized Tukey method for weighted multiple pairwise comparisons in an ANOVA model.

Examples

PGTukey(rep(6,4), Qx=2.42, ik=1, W=c(5,rep(1,5)) )

baolinwu/IPAM documentation built on May 9, 2019, 2:21 a.m.