eval.ROC: Evaluate ROC and AUC

In yunzhang813/simDeNet-R-Package: Simulation study on deconvolution for co-expression network

Description

This function performs ROC analysis for network estimation based on the ROCR package.

Usage

eval.ROC(est.str, true.str, plot.ROC=TRUE, show.AUC=TRUE, zoom=FALSE, lightPDF=FALSE, ...)

Arguments

est.str	symmetric matrix or array of symmetric matrices of estimated network structure, with elements taking values form 0 to 1 which measures the possibility of being an edge. It should have the same dimension as true.str.
true.str	symmetric matrix or array of symmetric matrices of true structure, usually with binary elements of 1 = edge and 0 = no edge. It should have the same dimension as est.str.
plot.ROC	logical value indicating if to plot the ROC curve. Default: TRUE.
show.AUC	logical value indicating if to show AUC on the plot. Default: TRUE.
zoom	logical value indicating if to zoom-in the plot. Default: FALSE.
lightPDF	logical value indicating if to subsample points for plotting in a PDF. Default: FALSE.
...	arguments passed to plot().

Value

A list with the following components:

perf	object of performence-class.
AUC	numeric value of the AUC statistic.
xval	maximum value plotted in the x-axis.
yval	maximum value plotted in the y-axis.

Examples

## Firstly, let's generate some data.
set.seed(999)
data("celltype")
mu.T <- expr[,ctab$Fastq_file_name[which(ctab$X3_letter_code=="ASM")]]
mu.N <- expr[,ctab$Fastq_file_name[which(ctab$X3_letter_code=="AEC")]]
## number of samples to simulate
n.samp <- 5
## parameters for correlation design of cell type T
rho <- c(0.9,0.8,0.7)
block.size <- c(5,10,15)
str.type <- c("interchangeable","decaying","star")
## one-step simulation
out.oneStepSim <- oneStepSim(n.samp, mu.T, mu.N, rho=rho, block.size=block.size, str.type=str.type)

## If we consider correlation network
est.str.pure <- abs(cor(t(out.oneStepSim$expr.pure.T)))
est.str.mixed <- abs(cor(t(out.oneStepSim$expr.mixed)))
## the true structure in cell type T should be
true.str <- out.oneStepSim$true.str.T

## ROC analysis
out.ROC.pure <- eval.ROC(est.str=est.str.pure, true.str=true.str, plot.ROC=TRUE, main="Pure samples")
out.ROC.pure$AUC
out.ROC.mixed <- eval.ROC(est.str=est.str.mixed, true.str=true.str, plot.ROC=TRUE, main="Mixed samples")
out.ROC.mixed$AUC

yunzhang813/simDeNet-R-Package documentation built on Dec. 24, 2019, 3:02 p.m.