cure.profile {logconcens}

Evaluate the Profile Log-Likelihood on a Grid of p_0-Values

Description

For each of a series of values for the cure parameter p_0 run the function logcon and evaluate the (normalized) log-likelihood at (\phi,p_0), where \phi is the log subprobability density returned by logcon. This serves for (approximate) joint likelihood maximization in (\phi,p_0).

Usage

cure.profile(x, p0grid=seq(0,0.95,0.05), knot.prec=IQR(x[x<Inf])/75,
                  reduce=TRUE, control=lc.control())

Arguments

x

a two-column matrix of n \geq 2 rows containing the data intervals.

p0grid

a vector of values p_0 for which the profile log-likelihood is to be evaluated.

knot.prec, reduce, control

arguments passed to the function logcon.

Value

A list containing the following values:

p0hat

the element in p0grid that maximizes the profile likelihood (in the very unlikely case of ties, only the smallest such element is returned).

status

the vector of (normalized) profile log-likelihood values for the elements of p0grid.

Note

For a large p0grid-vector (fine grid) computations may take a long time. Consider using the option adapt.p0 in the function logcon for a much faster method of joint likelihood maximization in (\phi,p_0).

Author(s)

Dominic Schuhmacher [e-mail]
Kaspar Rufibach [e-mail]
Lutz Duembgen [e-mail]

See Also

logcon, loglike

Examples

## The example from the logconcens-package help page:
set.seed(11)
x <- rgamma(50,3,1)
x <- cbind(x,ifelse(rexp(50,1/3) < x,Inf,x))

## Not run: 
plotint(x)
progrid <- seq(0.1,0.6,0.025)
prores <- cure.profile(x, progrid)
plot(progrid, prores$loglike)
prores$p0hat
res <- logcon(x, p0=prores$p0hat)
plot(res, type="survival")

## End(Not run)