family.rsm.object {marg}

Family Object for Regression-Scale Models

Description

Class of objects that characterize the error distribution of a regression-scale model.

Generation

This class of objects is returned by a call to a family.rsm generator function. See rsm.families for the distributions which are supported by the marg package. The object includes a list of functions and expressions that characterize the error distribution of a regression-scale model. These are used by the IRLS algorithm implemented in the rsm fitting routine. New families can be added to the ones already supported. See the demonstration file ‘margdemo.R’ that ships with the package. There is a print method for family.rsm objects which produces a simple summary without any detail; use unclass(family.rsm.object) to see the contents.

Structure

The following components, with the corresponding functionality, are required for a family.rsm object:

family

a character vector giving the family name.

g0

a function that yields minus the log density of the error distribution in the regression-scale model.

g1

a function that yields the first derivative of minus the log density.

g2

a function that yields the second derivative of minus the log density.

df

argument with NULL value; must be included to guarantee compatibility with the existing code.

k

argument with NULL value; must be included to guarantee compatibility with the existing code.

Note

For the sake of compatibility, the g0, g1 and g2 functions of a user-defined family can only take two arguments: y representing an observation and the ... argument which absorbes any additional arguments.

See Also

rsm.families, family.rsm, rsm