gGlobalAncova: Generalized GlobalAncova group test

In GlobalAncova: Global test for groups of variables via model comparisons

Description

Computation of a permutation test for the association between sets of variables (e.g. genes, SNPs, ...) and clinical entities. The variables can be continuous, binary, categorical, ordinal, or of mixed types. The test is carried out by comparing the deviances of the full generalized linear model and the reduced model lacking the design parameters of interest. The variable-wise models are summarized to a global test statistic for the complete set.

Usage

gGlobalAncova(data, formula.full, formula.red=~1, model.dat, Sets, sumstat=sum, perm=10000)

Arguments

data	data.frame of variables to be tested in sets (columns=variables); (multi-) categorical variables should be factors, ordinal variables should be ordered factors
formula.full	model formula for the full model
formula.red	model formula for the reduced model (that does not contain the terms of interest)
model.dat	data.frame of regressors, containing variables specified in formula.full and formula.red
Sets	vector of names or indices of variables or list of those, defining sets of variables
sumstat	function for summarizing univariate test statistics; default is sum
perm	number of permutations

Value

A data.frame with test statistic and p-value for each tested set.

Note

The test is fast for categorical data and categorical design variable. For other types of variables and more complex designs it is rather slow.

This work was supported by BMBF grant 01ZX1309B, Germany.

Author(s)

Reinhard Meister [e-mail]
Manuela Hummel [e-mail]

See Also

GlobalAncova

Examples

data(bindata)
gGlobalAncova(bindata[,-1], formula.full = ~group, model.dat = bindata, perm = 1000)

GlobalAncova documentation built on Nov. 8, 2020, 8:10 p.m.