coxph.tree {TimeVTree}

Function to Grow the Tree Using the Score Statistic

Description

This funtion finds the optimal cutpoints for the time-varying regression effects and grows the 'full tree' using the score statistic.

Usage

coxph.tree(survtime, survstatus, x, D = 3, method = "breslow", minfail = 10, 
            iter.max = 20, eps = 1e-06, type = 'mod')

Arguments

survtime

survival time/ follow up time of subjects

survstatus

survival status of subjects. 0 for censored and 1 for event

x

a data frame of covariates. In case of single covariate, use [,,drop =F] to keep the data frame structure

D

maximum depth the tree will grow. Default depth is 3.

method

argument for coxph function. Default is 'breslow'. See coxph for more details.

minfail

minimum number of unique events required in each block. Default is 10

iter.max

the maximum number of iteration in coxph; default is 20. See coxph for more details.

eps

argument for coxph function; default is 0.000001. See coxph for more details.

type

method to calculate the score statistic. Two options are available: 'mod' for the modified score statistic and 'ori' for the original score statistic. Default value is 'mod.' Modified score statistic is used in the bootstrap part

Details

coxph.tree takes in survival time, survival status, and covariates to grow the full tree. It follows one of the stopping rules: 1) when the pre-specified depth is reached, or 2) the number of events in a node is less than a prespecified number, or 3) the maximized score statistic is less than a default value (0.0001).

Currently, data need to be arranged in descending order of time and with no missing.

Value

coxph.tree returns an object of class 'coxphtree.'

The function output.coxphout is used to obtain and print a summary of the result.

An object of class 'coxphtree' is a list containing the following components:

D

Depth value specified in the argument

coef

coefficient values of predictors. First number represents depth and second number represents block number

lkl

Likelihood ratio value of each node

breakpt

Starting point of each node. Starting point of node at Depth= 0 to maximum Depth = D+1 is shown.

ntree

Number of cases in each node

nevent

Number of events in each node

nblocks

Number of blocks in each depth

nodes

Indicator that indicates whether the block was eligible for further split

nodetree

A table with depth, block, node, left right, maximum score, start time, end time, # of cases, and # of events

scoretest

Maximum score at each block

xnames

Name of predictors

failtime

The time when events occurred without duplicates

summary

coxph output of each block

pvalue

p-value to test validity of a change point against none

References

Xu, R. and Adak, S. (2002), Survival Analysis with Time-Varying Regression Effects Using a Tree-Based Approach. Biometrics, 58: 305-315.

Examples

##Call in alcohol data set
data('alcohol')
require(survival)

coxtree <- coxph.tree(alcohol[,'time'], alcohol[,'event'], 
                      x = alcohol[,'alc', drop = FALSE], D = 4)

nodetree <- output.coxphout(coxtree)

subtrees <- prune(nodetree)