Compute MLE for Censored Regression Models

DESCRIPTION:

Internal function used to compute estimates for censorReg .

USAGE:

censorReg.mlest(y, censor.codes, xmat, case.weights, distribution,
    link, initial=NULL, offset=0, parameter.fixed=rep(F,nparm),
    intercept=T, fixed=list(), control=NULL, YT.Y=NULL,
    YT.codes=NULL)

REQUIRED ARGUMENTS:

y
matrix of failure times or censoring intervals. Except for interval censored or counting data, this has only a single column. For interval censoring or counting data, the first column contains the start times, while the second column contains the end times for the interval.
censor.codes
vector of censor codes. These are: 1 exact failure 2 right censored 3 left censored 4 interval censored
xmat
the model design matrix.
case.weights
numeric vector of case weights (frequencies).
distribution
character string specifying the model distribution. This may be "extreme", "gaussian", or "logistic".
link
character string specifying the link function. This may be "log" or "identity".

OPTIONAL ARGUMENTS:

initial
numeric vector of initial values for the linear model parameters and scale. (The scale parameter is last.)
offset
numeric vector supplying the offset. This term is subtracted from the transformed (via link) response and truncations times prior to fitting the model.
parameter.fixed
a vector of boolean values. If the i-th value is TRUE, then the i-th parameter is fixed at the initial value.
intercept
a logical value, if TRUE an intercept term is included in the model.
fixed
the values of the fixed parameters.
control
the list object returned by censorReg.control. See censorReg.control for the components.
YT.Y
numeric matrix of truncation intervals, if any.
YT.codes
numeric vector of truncation codes, if any, with the same coding as censor.codes.

VALUE:

a list containing the results of fitting the model.

DETAILS:

A Fletcher-Powell algorithm is used in model fitting. The basic estimation code comes from W. Q. Meeker and is described in Meeker and Duke (1981) which details a program named CENSOR. Recently, W. Q. Meeker has refined the estimation algorithms, making them more general and stable (personal communication).

REFERENCES::

Meeker, W. Q. and Duke, S. D. (1981). CENSOR - A User-Oriented Computer Program for Life Data Analysis. The American Statistician, 35, 112.

SEE ALSO:

, .

EXAMPLES:

This is an internal routine that is not usually called by users.