metaSeq (version 1.12.0)

Stouffer.test: Stouffer's weighted Z-score method (Inverse normal method)

Description

Stouffer's method combines multiple weighted Z-scores which are calculated in each study. Although many weight can be introduced but weighting by sample-size is used in meta.oneside.noiseq.

Usage

Stouffer.test(pvals, na.mode = "notignore")

Arguments

pvals

A matrix coming from meta.oneside.noiseq function or other.oneside.pvalues, which is used for any one-sided p-values or probability.

na.mode

A string indicating how to treat NA in pvals. "notignore" means that genes having at least one NA is regarded as NA. "ignore" means NA is ignored and remaining data is used. By default, na.mode = "notignore".

References

Stouffer, S. A. and Suchman, E. A. and DeVinney, L. C. and Star, S. A. and Williams, R. M. Jr. (1949) The American Soldier, Vol. 1 - Adjustment during Army Life. Princeton, Princeton University Press.

Examples

data(BreastCancer)
library("snow")

# Experimental condition (1: BreastCancer, 0: Normal)
flag1 <- c(1,1,1,0,0, 1,0, 1,1,1,1,1,1,1,0, 1,1,0)

# Source of data
flag2 <- c("A","A","A","A","A", "B","B", "C","C","C","C","C","C","C","C", "D","D","D")

# readData function for meta-analysis
cds <- meta.readData(data = BreastCancer, factor = flag1, studies = flag2)

# oneside NOISeq for meta-analysis
# cl <- makeCluster(4, "SOCK")
# result <- meta.oneside.noiseq(cds, k = 0.5, norm = "tmm", replicates = "biological", factor = flag1, conditions = c(1, 0), studies = flag2, cl = cl)
# stopCluster(cl)

# Script above is very time-consumming step. Please use this pre-calculated result instead
data(Result.Meta)
result <- Result.Meta

# Fisher's method (without weighting)
F <- Fisher.test(result)
str(F)

# Stouffer's method (with weighting by sample-size)
S <- Stouffer.test(result)
str(S)