get.IDR {GMCM}

Posterior class probabilities, local, and adjusted IDRs.

Description

Functions for computing posterior cluster probabilities (get.prob) in the general GMCM as well as local and adjusted irreproducibility discovery rates (get.IDR) in the special GMCM.

Usage

get.IDR(x, par, threshold = 0.05, ...)

get.prob(x, theta, ...)

Arguments

x

A matrix of observations where rows corresponds to features and columns to studies.

par

A vector of length 4 where par[1] is mixture proportion of the irreproducible component, par[2] is the mean value, par[3] is the standard deviation, and par[4] is the correlation of the reproducible component.

threshold

The threshold level of the IDR rate.

...

Arguments passed to qgmm.marginal.

theta

A list of parameters for the full model as described in rtheta.

Value

get.IDR returns a list of length 5 with elements:

idr

A vector of the local idr values. I.e. the posterior probability that x[i, ] belongs to the irreproducible component.

IDR

A vector of the adjusted IDR values.

l

The number of reproducible features at the specified threshold.

threshold

The IDR threshold at which features are deemed reproducible.

Khat

A vector signifying whether the corresponding feature is reproducible or not.

get.prob returns a matrix where entry (i,j) is the posterior probability that the observation x[i, ] belongs to cluster j.

Note

From GMCM version 1.1 get.IDR has been an internal function. Use get.prop or get.IDR instead. The function can still be accessed with GMCM:::get.idr. get.idr returns a vector where the i'th entry is the posterior probability that observation i is irreproducible. It is a simple wrapper for get.prob.

Author(s)

Anders Ellern Bilgrau <[e-mail]>

References

Li, Q., Brown, J. B. J. B., Huang, H., & Bickel, P. J. (2011). Measuring reproducibility of high-throughput experiments. The Annals of Applied Statistics, 5(3), 1752-1779. doi:10.1214/11-AOAS466

Tewari, A., Giering, M., & Raghunathan, A. (2011). Parametric Characterization of Multimodal Distributions with Non-gaussian Modes. IEEE 11th International Conference on Data Mining Workshops, 2011, 286-292. doi:10.1109/ICDMW.2011.135

Examples

set.seed(1123)

# True parameters
true.par <- c(0.9, 2, 0.7, 0.6)

# Simulation of data from the GMCM model
data <-  SimulateGMCMData(n = 1000, par = true.par, d = 2)

# Initial parameters
init.par <- c(0.5, 1, 0.5, 0.9)

# Nelder-Mead optimization
nm.par   <- fit.meta.GMCM(data$u, init.par = init.par, method = "NM")

# Get IDR values
res <- get.IDR(data$u, nm.par, threshold = 0.05)

# Plot results
plot(data$u, col = res$Khat, pch = c(3,16)[data$K])