rocplot: Plot an ROC curve.

In rjbgoudie/structmcmc: Structural inference for Bayesian Networks and Variable Selection.

Description

Print a ROC curve given a variety of estimation methods.

Usage

  rocplot(true, maps, bnpmls, eps,
    thresholds = seq(from = 0, to = 1, by = 0.01),
    use.cpdags = F, verbose = F, ...)

Arguments

true	An object of class bn giving the true graph. This is converted to a CPDAG before comparision.
maps	An object of class bn.list. An estimate of the true graph, given by the MAP estimate given by the MCMC. This is internally converted to CPDAG.
bnpmls	An object of class bnpostmcmc.list. An object encapsulating a number of MCMC runs.
eps	An object of class ep.list.
thresholds	An object of class numeric, giving the thresholds at which to plot the graphs given by MCMC and PC-MCMC estimates
use.cpdags	A logical of length 1 indicating whether DAG should be converted to CPDAGs before computing the ROC curves.
verbose	A logical indicating whether to show progress bars etc.
...	Further arguments

Details

An alternative is given by: Werhli et al. Comparative evaluation of reverse engineering gene regulatory networks with relevance networks, graphical gaussian models and bayesian networks. Bioinformatics (2006) vol. 22 (20) pp. 2523

Value

A lattice object, containing the ROC plot

See Also

as.roc

Examples

x1 <- factor(c(1, 1, 0, 1, 1, 1))
x2 <- factor(c(0, 1, 0, 1, 0, 1))
x3 <- factor(c(0, 1, 0, 0, 0, 0))
x <- data.frame(x1 = x1, x2 = x2, x3 = x3)

exact <- posterior(x, "exact")
eppost <- ep(exact)
mcmc <- posterior(x, "mc3", nSamples = 1000, nBurnin = 100)

rocplot(true = bn(integer(0), c(1,3), integer(0)),
        eps = ep.list(eppost))
rocplot(true = bn(integer(0), integer(0), 2),
        eps = ep.list(eppost))

rjbgoudie/structmcmc documentation built on Nov. 3, 2020, 3:41 a.m.