bayesQR (version 2.2)

plot.bayesQR: Produce quantile plots or traceplots with plot.bayesQR

Description

plot.bayesQR is an S3 method that produces quantile plots, traceplots or posterior histograms based on the estimates obtained by the bayesQR function. For quantile plots, note that the more quantiles are estimated with bayesQR, the more detailed the quantile plot will be.

Usage

"plot"(x, var=NULL, quantile=NULL, burnin=0, credint=c(0.025,0.975), plottype=NULL, 
main=NULL, xlab=NULL, ylab=NULL, xlim=NULL, ylim=NULL, ...)

Arguments

x

an output object of the bayesQR function, i.e. an S3 object of class bayesQR.

var

vector containing the index or name of the variable(s) that has to be plotted (default=all variables).

quantile

vector containing the quantile(s) that has to be plotted (default=all quantiles).

burnin

the number of burnin draws that should be discared (default=0, meaning all draws are included).

credint

the width of the credible interval (default=c(0.025, 0.975)).

plottype

should be `quantile', `trace' or `hist'.

main

Main title of the plot (default="").

xlab

Label of the x-axis; if omitted, the value chosen based on the input data.

ylab

Label of the y-axis; if omitted, the value chosen based on the input data.

xlim

Plot region of the x-axis; if omitted, the value chosen based on the input data.

ylim

Plot region of the y-axis; if omitted, the value chosen based on the input data.

...

additional arguments that are passed to the generic plot function

Value

A (series of) quantile plot(s) or a (series of) traceplot(s).

Details

A quantile plot shows how the value of the regression parameter changes over a range of quantiles together with the associated credible interval. Note that the more quantiles are estimated, the more detailed the quantile plot will be. The minimum number of quantiles to plot is 2. A posterior histogram provides a graphical representation of the marginal posterior distribution of the regression parameters. A traceplot gives the evolution of the MCMC chain and can be used as graphical check of convergence. Note that more formal checks of convergence exist (see, for example, the coda package).

Examples

# Simulate data from heteroskedastic regression
set.seed(66)
n <- 200
X <- runif(n=n,min=0,max=10)
X <- X
y <- 1 + 2*X + rnorm(n=n, mean=0, sd=.6*X)

# Analyze 5 quantiles using default prior
out <- bayesQR(y ~ X, quantile=c(.05,.25,.5,.75,.95), ndraw=5000)

# Check traceplot of first variable of .75 quantile regression 
plot(out, var=1, quantile=.75, plottype="trace")

# Check posterior histogram of first variable of .5 quantile regression 
plot(out, var=1, quantile=.5, plottype="hist")

# Create default quantile plot of first variable
plot(out, var=1, plottype="quantile")

# Create quantile plot of second variable with 90% credible interval
plot(out, var="X", credint=c(.05, .95), plottype="quantile", main="This is an example")