LFDR.MME (version 1.0)

LFDR.MM: Performs a Multiple Hypothesis Testing Using the Method of Moments

Description

Based on a given vector of chi-square test statistics, provides estimates of local false discoveries.

Usage

LFDR.MM(x)

Arguments

x

A vector of chi-square test statistics with one degree of freedom.

Value

Outputs three elements as seen below:

pi0.hat

estimate of proportion of unassocaited features \(\pi_0\).

ncp.hat

estimate of the non-centrality parameter \(\lambda\) of the chi-square model for associated features.

lfdr.hat

estimates of local false discovery rates.

Details

For \(N\) given features (genes, proteins, SNPs, etc.), the function tests the null hypothesis \(H_{0i}\), \(i=1,\ldots,N\), indicating that there is no association between feature \(i\) and a specific disease, versus its alternative hypothesis \(H_{1i}\). For each unassociated feature \(i\), it is suppoed that the corresponding test stiatistic \(x_i\) follows a central chi-square distribution with one degree of freedom. For each associated feature \(i\), it is assumed that the corresponding test stiatistic \(x_i\) follows a non-central chi-square distribution with one degree of freedom and non-centrality parameter \(\lambda\). In this packag, association is measured by estimating the local false discovery rate (LFDR), the posterior probability that the null hypothesis \(H_{0i}\) given the test statistic \(x_i\) is true. This package returns three components as mentioned in the Value section.

References

Karimnezhad, A. (2020). A Simple Yet Efficient Parametric Method of Local False Discovery Rate Estimation Designed for Genome-Wide Association Data Analysis. Retrieved from https://arxiv.org/abs/1909.13307

Examples

# NOT RUN {
# vector of test statistics for assocaited features
stat.assoc<- rchisq(n=1000,df=1, ncp = 3)

# vector of test statistics for unassocaited features
stat.unassoc<- rchisq(n=9000,df=1, ncp = 0)

# vector of test statistics
stat<- c(stat.assoc,stat.unassoc)

output <- LFDR.MM(x=stat)

# Estimated pi0
output$p0.hat

# Estimated non-centrality parameter
output$ncp.hat

# Estimated LFDRs
output$lfdr.hat
# }