BioMark (version 0.4.5)

HCthresh: Biomarker thresholding by Higher Criticism

Description

Higher Criticism (HC) is a second-level significance testing approach to determine which variables in a multivariate set show significant differences in two classes. Function HCthresh selects those p values that are significantly different from what would be expected from their uniform distribution under the null hypothesis.

Usage

HCthresh(pvec, alpha = 0.1, plotit = FALSE)

Arguments

pvec

Vector of p values.

alpha

Parameter of the HC approach: the maximal fraction of differentially expressed p values.

plotit

Logical, whether or not a plot should be produced.

Value

indices of the p values satisfying the HC criterion.

Details

In HC, one tests the deviation of the expected behaviour of p values under a null distribution. Function HCthresh implements the approach by Donoho and Jin to find out which of these correspond to real differences. The prerequisites are that the true biomarkers are rare (consist of only a small fraction of all variables) and weak (are not able to discriminate between the two classes all by themselves).

References

David Donoho and Jiashun Jin: Higher criticism thresholding: Optimal feature selection when useful features are rare and weak. PNAS 108:14790-14795 (2008).

Ron Wehrens and Pietro Franceschi: Thresholding for Biomarker Selection in Multivariate Data using Higher Criticism. Mol. Biosystems (2012). In press. DOI: 10.1039/C2MB25121C

See Also

get.biom for general approaches to obtain biomarkers based on multivariate discriminant methods and t statistics

Examples

data(spikedApples)
bms <- get.biom(spikedApples$dataMatrix, rep(0:1, each = 10),
                type = "coef", fmethod = "studentt")
bms.pvalues <- 2 * (1 - pt(abs(bms[[1]]), 18))
sum(bms.pvalues < .05)                           ## 15
sum(p.adjust(bms.pvalues, method = "fdr") < .05) ## 4
signif.bms <- HCthresh(bms.pvalues, plotit = TRUE)
length(signif.bms)                               ## 11