Functions for visualizing the output from bmonomvn,
particularly the posterior standard deviation estimates of the mean
vector and covariance matrix, and samples from the solution to a
Quadratic Program
# S3 method for monomvn
plot(x, which=c("mu", "S", "Snz", "Sinz", "QP"),
xaxis=c("numna", "index"), main=NULL, uselog=FALSE, ...)a "monomvn"-class object that must be named x
for generic plotting
determines the parameter whose standard deviation
to be visualized: the mean vector ("mu" for sqrt($mu.var));
the covariance matrix ("S" for sqrt($S.var)),
or "S{i}nz" for sqrt($S{i}.nz), which both result in
an image plot; or the distribution of solutions
$W to a Quadratic Program that may be obtained by
supplying QP = TRUE as input to bmonomvn
indicates how x-axis (or x- and y-axis in the case
of which = "S" || "S{i}nz") should be displayed. The default option
"numna" shows the (ordered) number of missing entries
(NAs) in the corresponding column, whereas "index"
simply uses the column index; see details below
optional text to be added to the main title of the plots;
the default of NULL causes the automatic generation of a
title
a logical which, when TRUE, causes the log
of the standard deviation to be plotted instead
passed to plot.default
The only output of this function is beautiful plots
Currently, this function only provides a visualization of the
posterior standard deviation estimates of the parameters, and
the distributions of samples from the posterior of the solution
to a specified Quadratic Program. Therefore
it only works on the output from bmonomvn
All types of visualization (specified by which) are presented
in the order of the number of missing entries in the columns of the
data passed as input to bmonomvn.
In the case of which = "mu"
this means that y-values are presented in the order x$o, where
the x-axis is either 1:length(x$o) in the case of
xaxis = "index", or x$na[x$o] in the case of xaxis
= "numna". When which = "S" is given the resulting
image plot is likewise ordered by x$o where the
x- and y-axis are as above, except that in the case
where xaxis = "numna" the repeated counts of NAs are
are adjusted by small increments so that x and y arguments to
image are distinct. Since a boxplot is
used when which = "QP" it may be that xaxis = "index"
is preferred