Plot the MAP Tree for each height encountered by the Markov Chain

Description

Plot the maximum a' posteriori (MAP) tree as a function of tree height, and show the log posterior probabilities for comparison.

Usage

1
tgp.trees(out, heights = NULL, main = NULL, ...)

Arguments

out

"tgp"-class object which is the output of one the model functions with tree support (e.g. btgpllm)

heights

Index vector of length less than length(out$trees) describing trees to plot by their height. Default (NULL) is to plot all trees, one for each height encountered when sampling from the Markov chain of the tree posterior. This is equivalent to heights = out$posts$height. Specifying heights = "map" causes (only) the maximum a' posteriori (MAP) height tree to be plotted

main

Optional character string to add to the main title of the plot

...

Extra arguments to the draw.tree function from maptree

Details

The maximum a' posteriori (MAP) tree encountered at each height (in the MCMC chain) is plotted, and the log posterior probabilities are shown for comparison. The text at the branches in the tree show the splitting variable and value. The text at the leaves show the number of input data points (X and Z) that fall into the region(s) along with an estimate of the variability therein.

Value

The only output of this function is beautiful tree diagrams.

Note

Plotting trees that the maptree library is installed, which itself requires that the combinat library also be installed.

See vignette("tgp") and the examples sections of the functions under “See Also”, below

Author(s)

Robert B. Gramacy, rbgramacy@chicagobooth.edu, and Matt Taddy, taddy@chicagobooth.edu

References

http://bobby.gramacy.com/r_packages/tgp

See Also

bgpllm, btlm, blm, bgp, btgpllm, plot.tgp, mapT, vignette("tgp")