Combine p-values

DESCRIPTION:

Nonparametric combination of multivariate p-values in permutation testing

USAGE:

combinePValues.Fisher(p, ...) 
combinePValues.Liptak(p, ...) 
combinePValues.Tippett(p, ...) 
combinePValues.chisquare(p, ...) 
combinePValues.halfnormal(p, ...) 

REQUIRED ARGUMENTS:

p
a matrix containing p-values; each column corresponds to a variable (these need not be independent) and each row to a resample (typically a set of p-values computed from a randomly-permuted version of the original data).

OPTIONAL ARGUMENTS:

...
other arguments are currently ignored.

VALUE:

vector, length equal to the number of rows of p. The ith value depends only on the ith row of p.

DETAILS:

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].

REFERENCES:

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.

SEE ALSO:

, .

EXAMPLES:

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)