LaplacesDemon (version 16.1.0)

plotMatrix: Plot a Numerical Matrix

Description

This function plots a numerical matrix, and is often used to plot the following matrices: correlation, covariance, distance, and precision.

Usage

plotMatrix(x, col=colorRampPalette(c("red","black","green"))(100),
     cex=1, circle=TRUE, order=FALSE, zlim=NULL, title="", PDF=FALSE, …)

Arguments

x

This required argument is a numerical matrix, or an object of class bayesfactor, demonoid, iterquad, laplace, pmc, posteriorchecks, or vb. See more information below regarding these classes. One component of a blocked proposal covariance matrix must be pointed to explicitly, rather than to the object of class demonoid.

col

This argument specifies the colors of the circles. By default, the colorRampPalette function colors strong positive correlation as green, zero correlation as black, and strong negative correlation as red, and provides 100 color gradations.

cex

When circle=TRUE, this argument specifies the size of the marginal text, the names of the parameters or variables, and defaults to 1.

circle

Logical. When TRUE, each element in the numeric matrix is represented with a circle, and a larger circle is assigned to elements that are farther from zero. Also, when TRUE, the gradation scale does not appear to the right of the plot.

order

Logical. This argument defaults to FALSE, and presents the parameters or variables in the same order as in the numeric matrix. When TRUE, the parameters or variables are ordered using principal components analysis.

zlim

When circle=FALSE, the gradation scale may be constrained to an interval by zlim, such as zlim=c(-1,1), and only values within the interval are plotted.

title

This argument specifies the title of the plot, and the default does not include a title. When x is of class posteriorchecks, the title is changed to Posterior Correlation.

PDF

Logical. When TRUE, the plot is saved as a .pdf file.

…

Additional arguments are unused.

Details

The plotMatrix function produces one of two styles of plots, depending on the circle argument. A \(K \times K\) numeric matrix of \(K\) parameters or variables is plotted. The plot is a matrix of the same dimensions, in which each element is colored (and sized, when circle=TRUE) according to its value.

Although plotMatrix does not provide the same detail as a numeric matrix, it is easier to discover elements of interest according to color (and size when circle=TRUE).

The plotMatrix function is not inherently Bayesian, and does not include uncertainty in matrices. Nonetheless, it is included because it is a useful graphical presentation of a numeric matrices, and is recommended to be used with the posterior correlation matrix in an object of class posteriorchecks.

When x is an object of class bayesfactor, matrix B is plotted. When x is an object of class demonoid (if it is a matrix), iterquad, laplace, pmc, or vb, the covariance matrix Covar is plotted. When x is an object of class posteriorchecks, the posterior correlation matrix is plotted.

This is a modified version of the circle.corr function of Taiyun Wei.

See Also

PosteriorChecks

Examples

# NOT RUN {
library(LaplacesDemon)
### Although it is most commonly used with an object of class
### posteriorchecks, it is applied here to a different correlation matrix.
data(mtcars)
plotMatrix(cor(mtcars), col=colorRampPalette(c("green","gray10","red"))(100),
     cex=1, circle=FALSE, order=TRUE)
plotMatrix(cor(mtcars), col=colorRampPalette(c("green","gray10","red"))(100),
     cex=1, circle=TRUE, order=TRUE)
# }