pscl (version 1.4.6)

tracex: trace plot of MCMC iterates, posterior density of legislators' ideal points

Description

Produces a trace plot of the MCMC samples from the posterior density of legislators' ideal points.

Usage

tracex(object, legis=NULL, d=1, conf.int=0.95,
       multi = FALSE, burnin=NULL,span=.25,
       legendLoc="topright")

Arguments

object

an object of class ideal.

legis

a vector of either the names of legislators (or partial matches of the names as given in the dimnames of object$x.

d

numeric, either a scalar or a vector of length two, the dimension(s) to be traced.

conf.int

numeric, the level of the confidence interval on the posterior mean to be plotted.

multi

logical, multiple plotting panels, one per legislators? If FALSE (default) and length(d)==2, display traces for all selected legislators' ideal points on the one plot.

burnin

of the recorded MCMC samples, how many to discard as burnin? Default is NULL, in which case the value of burnin in the ideal object is used.

span

numeric, a proportion, the span to be used when calling loess to generate a moving average for trace plots when d=1

legendLoc

numeric or character, and possibly a vector, specifying where to place the legend when d=1; setting legendLoc=NULL will suppress the legend for all requested trace plots

Details

Produces a trace plot showing the history of the MCMC iterations for the ideal point of each of the legislators (partially) named in legis. For d=1, each trace plot includes a trace over iterations, the cumulative mean, a moving average, the MCMC-based estimate of the mean of the posterior, and a confidence interval (specified by conf.int) around the mean of the posterior (using the estimated quantiles) of the respective MCMC iterates). All of these values are calculated discarding the initial burnin iterations.

When d is a vector of length two, a 2-dimensional trace plot is displayed, with the d[1] dimension on the horizontal axis, and the d[2] dimension on the vertical axis.

When d=1, a legend will be placed on the plot; the option legendLoc controls the placing of the legend. legendLoc may be a vector, specifying a unique legend location for each requested trace plots. If legendLoc is of length 1, it will be replicated to have length equal to the number of requested trace plots.

See Also

ideal; pmatch for matching legislators' names. See legend for valid options to legendLoc.

Examples

data(s109)
## short run for demo purposes only
id1 <- ideal(s109,
             normalize=TRUE,
             maxiter=500,
             burnin=100,thin=10,
             verbose=TRUE)
tracex(id1,legis="KENN")

## n.b., no such legislator named Thomas Bayes
tracex(id1,legis=c("KENN","BOX","KYL","Thomas Bayes"))

id2 <- ideal(s109,
             d=2,             ## unidentified!
             maxiter=5000,   
             burnin=0,
             thin=50)
tracex(id2,d=1,legis=c("KENNEDY","BOXER","KYL","Thomas Bayes"))
tracex(id2,d=2,legis=c("KENNEDY","BOXER","KYL","Thomas Bayes"))
tracex(id2,d=1:2,
       legis=c("KENNEDY","BOXER","KYL","Thomas Bayes"))

## partial matching
tracex(id2,d=1:2,
       legis=c("KENN","BOX","BID","SNO","SPEC","MCCA","KYL",
               "Thomas Bayes"),
       multi=TRUE)