mlreg {statmod}R Documentation

Fit a Linear Model by Maximum Likelihood

Description

Fits a linear model by maximum likelihood assuming a variety of response distributions.

Usage

mlreg.fit(X, y, weights=NULL, dist="logistic", init=NULL, scale=NULL)
mlreg.fit.zero(y, weights=NULL, dist="logistic", init=NULL, scale=NULL)

Arguments

X the design matrix. Rows containing missing values (in X or y) will be removed.
y numeric response vector. Missing values will be removed.
weights vector of non-negative weights.
dist character string giving the name of the response distribution. The possibilities are "extreme", "logistic", "gaussian", "weibull", "exponential", "rayleigh", "loggaussian", "lognormal", "loglogistic" or "t" on 4 df.
init numeric vector of initial values for the parameters.
scale if specified then the scale parameter is fixed at the given value.

Details

This function is merely a convenient wrapper for calling the survreg.fit function, which is part of the survival library by Terry Therneau. It fits the model y = X*b + scale*e where b is the vector of regression coefficients and e is a vector of mean-zero errors, by maximum likelihood.

The function mlreg.fit.zero assumes that the mean is zero and fits y = scale*e, estimating only the scale parameter.

Value

See the documentation for survreg.object

Author(s)

Gordon Smyth

See Also

survreg, survreg.object

Examples

x <- 1:50
y <- x + 2*rnorm(50)
X <- cbind(1,x)
out <- mlreg.fit(X,y,dist="logistic")

[Package Contents]