Learn R Programming

BayesTreePrior (version 1.0.1)

Bayesian Tree Prior Simulation

Description

Provides a way to simulate from the prior distribution of Bayesian trees by Chipman et al. (1998) . The prior distribution of Bayesian trees is highly dependent on the design matrix X, therefore using the suggested hyperparameters by Chipman et al. (1998) is not recommended and could lead to unexpected prior distribution. This work is part of my master thesis (expected 2016).

Copy Link

Version

Install

install.packages('BayesTreePrior')

Monthly Downloads

194

Version

1.0.1

License

GPL-3

Maintainer

Alexia Jolicoeur-Martineau

Last Published

July 4th, 2016

Functions in BayesTreePrior (1.0.1)

GetListUniqueSplits

Unique splits that leads to children with more than $minpart$ nodes.
NumBotMaxDepthX

Number of bottom nodes and depth in the general case (Case #4).
E_alpha

Expected value of the number of bottom nodes in the unrealistic case where we assume that the number of variables and possible splits are infinite (therefore P(T) is not dependent on the design matrix X) and $\beta=0$ (Case #1).
BayesTreePriorOrthogonal

Simulation of the tree prior in the case where we have one single variable (Case #3).
BayesTreePriorOrthogonalInf

Simulation of the tree prior in the unrealistic case where we assume that the number of variables and possible splits are infinite (therefore P(T) is not dependent on the design matrix X) (Case #2).
BayesTreePriorNotOrthogonal

Simulation of the tree prior in the general case (Case #4).
p_split

Probability of split of the tree prior.
NumBotMaxDepth

Number of bottom nodes and depth in the case where we have one single variable (Case #3).
NumBotMaxDepth_inf

Number of bottom nodes and depth in the unrealistic case where we assume that the number of variables and possible splits are infinite (therefore P(T) is not dependent on the design matrix X) (Case #2).
BayesTreePrior

Simulation of the tree prior.
Var_alpha

Variance of the number of bottom nodes in the unrealistic case where we assume that the number of variables and possible splits are infinite (therefore P(T) is not dependent on the design matrix X) and $\beta=0$ (Case #1).