mertree: Regression Trees for Longitudinal and Clustered Data

In bgreenwell/mertree: Mixed-Effects Regression Trees for Longitudinal and Clustered Data

Description

Fits an unbiased regression tree to longitudinal or clustered data by iterating back and forth between a conditional inference regression tree to capture complex interactions and nonlinear relationaships and a linear mixed-effects model to capture complex correlation structure.

Usage

mertree(formula, data, unbiased = TRUE, initial_re, REML = TRUE,
  lmer.control = lmerControl(calc.derivs = FALSE), lmer.verbose = 0L,
  tree.control = if (unbiased) {     ctree_control() } else {    
  rpart.control() }, cv = TRUE, tol = 0.001, maxiter = 100L,
  do.trace = FALSE)

Arguments

formula	An appropriate lmer-style formula.
data	An optional data frame containing the variables named in formula.
unbiased	Logical indicating whether or not to use a conditional inference tree. Default is TRUE.
initial_re	Numeric vector containing the initial values for the random effects. If omitted then defaults to zero.
REML	Logical indicating whether or not the estimates should be chosen to optimize the REML criterion (as opposed to the log-likelihood).
lmer.control	List of control parameters for lmer.
lmer.verbose	Integer specifying the verbosity of output printed during the call to lmer. If > 0, verbose output is generated during the optimization of the parameter estimates. If > 1, verbose output is generated during the individual PIRLS steps. Default is 0L meaning to supress such output.
tree.control	List of control parameters for ctree or rpart.
cv	Logical indicating whether or not to prune each tree based on cross-validations. Only used when unbiased = FALSE. Default is TRUE.
tol	The desired accuracy (convergence tolerance). Default is 0.001)
maxiter	Integer specifying the maximum number of iterations. Default is 1000.
do.trace	Logical indicating whether or not to print trace information.

bgreenwell/mertree documentation built on May 12, 2019, 8:19 p.m.