CoxBoost (version 1.5.1)

optimCoxBoostPenalty: Coarse line search for adequate penalty parameter

Description

This routine helps in finding a penalty value that leads to an ``optimal'' number of boosting steps for CoxBoost, determined by cross-validation, that is not too small/in a specified range.

Usage

optimCoxBoostPenalty(
  time,
  status,
  x,
  minstepno = 50,
  maxstepno = 200,
  start.penalty = 9 * sum(status == 1),
  iter.max = 10,
  upper.margin = 0.05,
  parallel = FALSE,
  trace = FALSE,
  ...
)

Value

List with element penalty containing the ``optimal'' penalty and cv.res containing the corresponding result of cv.CoxBoost.

Arguments

time

vector of length n specifying the observed times.

status

censoring indicator, i.e., vector of length n with entries 0 for censored observations and 1 for uncensored observations. If this vector contains elements not equal to 0 or 1, these are taken to indicate events from a competing risk and a model for the subdistribution hazard with respect to event 1 is fitted (see e.g. Fine and Gray, 1999).

x

n * p matrix of covariates.

minstepno, maxstepno

range of boosting steps in which the ``optimal'' number of boosting steps is wanted to be.

start.penalty

start value for the search for the appropriate penalty.

iter.max

maximum number of search iterations.

upper.margin

specifies the fraction of maxstepno which is used as an upper margin in which a cross-validation minimum is not taken to be one. This is necessary because of random fluctuations of cross-validated partial log-likelihood.

parallel

logical value indicating whether computations in the cross-validation folds should be performed in parallel on a compute cluster. Parallelization is performed via the package snowfall and the initialization function of of this package, sfInit, should be called before calling cv.CoxBoost.

trace

logical value indicating whether information on progress should be printed.

...

miscellaneous parameters for cv.CoxBoost.

Author

Written by Harald Binder .

Details

The penalty parameter for CoxBoost has to be chosen only very coarsely. In Tutz and Binder (2006) it is suggested for likelihood based boosting just to make sure, that the optimal number of boosting steps, according to some criterion such as cross-validation, is larger or equal to 50. With a smaller number of steps, boosting may become too ``greedy'' and show sub-optimal performance. This procedure uses a very coarse line search and so one should specify a rather large range of boosting steps.

References

Tutz, G. and Binder, H. (2006) Generalized additive modelling with implicit variable selection by likelihood based boosting. Biometrics, 62:961-971.

See Also

CoxBoost, cv.CoxBoost

Examples

# \donttest{
#   Generate some survival data with 10 informative covariates
n <- 200; p <- 100
beta <- c(rep(1,10),rep(0,p-10))
x <- matrix(rnorm(n*p),n,p)
real.time <- -(log(runif(n)))/(10*exp(drop(x %*% beta)))
cens.time <- rexp(n,rate=1/10)
status <- ifelse(real.time <= cens.time,1,0)
obs.time <- ifelse(real.time <= cens.time,real.time,cens.time)

#  determine penalty parameter

optim.res <- optimCoxBoostPenalty(time=obs.time,status=status,x=x,
                                  trace=TRUE,start.penalty=500)

#   Fit with obtained penalty parameter and optimal number of boosting
#   steps obtained by cross-validation

cbfit <- CoxBoost(time=obs.time,status=status,x=x,
                  stepno=optim.res$cv.res$optimal.step,
                  penalty=optim.res$penalty)
summary(cbfit)

# }