getDensityMatrix: Calculate density matrix from raw p-value matrix

In nem: (Dynamic) Nested Effects Models and Deterministic Effects Propagation Networks to reconstruct phenotypic hierarchies

Description

Fit a 3 component BUM model to each column of a raw p-value matrix.

Usage

	getDensityMatrix(Porig, dirname=NULL, startab=c(0.3,10), startlam=c(0.6,0.1,0.3), tol=1e-4)

Arguments

Porig	matrix of raw p-values
dirname	name of a directory to save histograms and QQ-plots to. If dirname=NULL, then the plots are made to the screen, and after each fit the user is asked to press a key in order to continue.
startab	start values for alpha and beta parameter
startlam	start values for mixing coefficients
tol	convergence tolerance: If the absolute likelihood ratio -1 becomes smaller than this value, then the EM algorithm is supposed to be converged.

Details

The BUM density model consists of 3 components: f(x) = lambda_1 + lambda_2*dbeta(x,alpha,1) + lambda_3*dbeta(x,1,beta). The mixing coefficients and the parameters alpha and beta are fitted together via an EM algorithm.

Value

log-density matrix of same dimensions as Porig: The log-densities can be interpreted as log signal-to-noise ratios. A value > 0 means higher signal than noise, and a value < 0 a higher noise than signal.

Note

Note the difference to the previous package version: the LOG-density is returned now!

Author(s)

Holger Froehlich

nem documentation built on Oct. 31, 2019, 2:12 a.m.