plot.BayesSUR {BayesSUR}R Documentation

create a selection of plots for a BayesSUR class object

Description

plot method for class BayesSUR. This is the main plot function to be called by the user. This function calls one or several of the following functions: plotEstimator(), plotGraph(), plotMCMCdiag(), plotManhattan(), plotNetwork(), plotCPO().

Usage

## S3 method for class 'BayesSUR'
plot(x, estimator = NULL, type = NULL, ...)

Arguments

x

an object of class BayesSUR

estimator

It is in c(NULL, 'beta', 'gamma', 'Gy', 'logP', 'CPO') and works by combining with argument type.

  • If estimator is in c("beta", "gamma", "Gy") and argument type="heatmap", it prints heatmaps of the specified estimator in estimator by a call to to function plotEstimator() for more other arguments.

  • If estimator="Gy" and argument type="graph", it prints a structure graph of "Gy" by a call to function plotGraph() for more other arguments.

  • If estimator=c("gamma", "Gy") and argument type="network", it prints the estimated network between the response variables and predictors with nonzero coefficients by a call to function plotMCMCdiag() for more other arguments.

  • If estimator=NULL (default) and type=NULL (default), it interactively prints the plots of estimators (i.e., beta, gamma and (or) Gy), response graph Gy, network, Manhattan and MCMC diagnostics.

type

It is one of NULL, "heatmap", "graph", "network", "Manhattan" and "diagnostics", and works by combining with argument estimator.

  • If type="Manhattan" and argument estimator="gamma", it prints Manhattan-like plots for marginal posterior inclusion probabilities (mPIP) and numbers of associated response variables for individual predictors by a call to function plotManhattan() for more other arguments.

  • If type="diagnostics" and argument estimator="logP" it shows trace plots and diagnostic density plots of a fitted model by a call to function plotMCMCdiag() for more other arguments.

  • If type="diagnostics" and argument estimator="CPO", it shows the conditional predictive ordinate (CPO) for each individual of a fitted model by a call to function plotCPO() for more other arguments.

...

other arguments, see functions plotEstimator(), plotGraph(), plotNetwork(), plotManhattan(), plotMCMCdiag() or plotCPO()

Examples

data("exampleEQTL", package = "BayesSUR")
hyperpar = list( a_w = 2 , b_w = 5 )

set.seed(9173)
fit <- BayesSUR(Y = exampleEQTL[["blockList"]][[1]], 
                X = exampleEQTL[["blockList"]][[2]],
                data = exampleEQTL[["data"]], outFilePath = tempdir(),
                nIter = 100, burnin = 0, nChains = 2, gammaPrior = "hotspot",
                hyperpar = hyperpar, tmpFolder = "tmp/" )

## check output
## Not run: 
# Show the interactive plots. Note that it needs at least 2000*(nbloc+1) iterations 
# for the diagnostic plots where nbloc=3 by default 
plot(fit)

## End(Not run)

## plot heatmaps of the estimated beta, gamma and Gy
plot(fit, estimator=c("beta", "gamma", "Gy"), type="heatmap")

## plot estimated graph of responses Gy
plot(fit, estimator="Gy", type="graph")

## plot network between response variables and associated predictors
plot(fit, estimator=c("gamma", "Gy"), type="network")

## print Manhattan-like plots
plot(fit, estimator="gamma", type="Manhattan")

## print MCMC diagnostic plots
plot(fit, estimator="logP", type="diagnostics")


[Package BayesSUR version 2.0-1 Index]