combinePValues.Fisher(p, ...) combinePValues.Liptak(p, ...) combinePValues.Tippett(p, ...) combinePValues.chisquare(p, ...) combinePValues.halfnormal(p, ...)
p
.
The
i
th value depends only on the
i
th row of
p
.
A combination function should be decreasing (non-increasing) in each argument
and should approach its maximum value (possibly
Inf
) as any argument
approaches 0. In other words, when a value is more significant
(
p
-values closer to 0) the result is larger.
The result need not be in the range
[0,1]
.
Pesarin, F. (2001),
Multivariate Permutation Tests with Applications to
Biostatistics: Nonparametric Combination Methodology ,
Wiley, Chichester, UK.
We wish to thank Dr. Luigi Salmaso for help designing and testing
these functions.
set.seed(0) x <- matrix(rnorm(15*3), 15) treatment <- rep(1:2, length=15) result <- permutationTestMeans(x, treatment = treatment, seed=1, combine = list(1:3, 1:2)) result update(result, combinationFunction = combinePValues.Liptak)