EBcoexpress (version 1.16.0)

priorDiagnostic: Visual diagnostic for the EBcoexpress prior

Description

A visual diagnostic used to check the prior estimated by an ebCoexpressSeries function against the data using the prior predictive distribution specified by the EBcoexpress model. The function compares the empirical prior predictive distribution of the (transformed) correlations in one condition against the theoretical prior predictive distribution

Usage

priorDiagnostic(D, conditions, ebOutObj, focusCond, seed = NULL, colx = "red", applyTransform = TRUE, subsize = NULL, ...)

Arguments

D

The correlation matrix output of makeMyD()

conditions

The conditions array

ebOutObj

The structured list output from an ebCoexpressSeries function

focusCond

A condition whose correlations will be used in the diagnostic. We suggest running the diagnostic for each condition, one at a time

seed

A seed for making the subsize subselection deterministic; has no effect if subsize= is left NULL

colx

A color for the fitted marginal distribution. Defaults to red

applyTransform

Should Fisher's Z-transformation be applied? Defaults to TRUE

subsize

If non-NULL, a value less than the 1st dimension of D (p). The diagnostic will use subsize randomly chosen correlations from the condition in its computation of the empirical density instead of all p pairs; by default, all pairs are used. We suggest use of this option when the number of pairs is very large

...

Other parameters to be passed to plot()

Value

Returns invisible(NULL)

Details

This function is a diagnostic tool for checking the prior distribution selected by the EM during an ebCoexpressSeries function's computations using the prior predictive distribution. The better the prior fits the observed data, the more confidence we should have in the posterior probabilities generated by the EM.

When run, the user specifies a condition. All of the (transformed) correlations from that condition (or just some of them if the subsize= option is non-NULL) will be used to estimate the empirical prior predictive distribution of the data in that condition; this will be plotted in black. The diagnostic then calculates the the theoretical prior predictive distribution and plots it using a dashed, colored line (set by colx=). If the two densities are similar, this indicates the selected prior fits the data in this condition well. The process can and should be repeated for all other conditions

References

Dawson JA and Kendziorski C. An empirical Bayesian approach for identifying differential co-expression in high-throughput experiments. (2011) Biometrics. E-publication before print: http://onlinelibrary.wiley.com/doi/10.1111/j.1541-0420.2011.01688.x/abstract

Examples

data(fiftyGenes)
tinyCond <- c(rep(1,100),rep(2,25))
tinyPat <- ebPatterns(c("1,1","1,2"))
D <- makeMyD(fiftyGenes, tinyCond, useBWMC=TRUE)
set.seed(3)
initHP <- initializeHP(D, tinyCond)

zout <- ebCoexpressZeroStep(D, tinyCond, tinyPat, initHP)
par(mfrow=c(2,1))
priorDiagnostic(D, tinyCond, zout, 1)
priorDiagnostic(D, tinyCond, zout, 2)
par(mfrow=c(1,1))