perm_test: perform permutation test
In floatofmath/adaperm: Permutation tests for adaptive designs

Description Usage Arguments Details Value Author(s)

Perform a permutation test (stratified by stages)

1
2
3

perm_test(x1, x2, g1, g2, stat, B, x3 = NULL, g3 = NULL, restricted,
  type = c("non-randomized", "randomized", "midp", "davison_hinkley"),
  stratified = TRUE)

`x1`	First stage observations
`x2`	Second stage observations
`g1`	First stage group assignments
`g2`	Second stage group assignments
`stat`	Function that computes test statistic
`B`	Number of permutations to use
`x3`	Third stage observations
`g3`	Third stage group assignments
`restricted`	Should group sizes be considered fixed
`type`	Type of p-value to compute (see details)
`stratified`	should permutation be stratified by stage

type specifies which permutation p-value to use. "non-randomized" will use the p-value of the non-randomized permutation test. "randomized" will use a randomized p-value that subtracts a random quantity from the non-randomized p-value, such that the a test that rejects for p <= alhpa has size exactly alpha. "midp" computes the mid-p-value which is the expected value of the randomized p-value. Davison-Hinkley returns the fraction between the permutation test statistics larger than t plus one and the permutation test statistics different from t plus one. The latter is mostly usefull when not all possible permutations are used.

p-value of the permutation test

Florian Klinglmueller

floatofmath/adaperm documentation built on May 16, 2019, 1:18 p.m.