vcov.mppm: Calculate Variance-Covariance Matrix for Fitted Multiple Point Process Model

Description

Given a fitted multiple point process model, calculate the variance-covariance matrix of the parameter estimates.

Usage

## S3 method for class 'mppm':
vcov(object, ..., what="vcov", err="fatal")

Arguments

object

A multiple point process model (object of class "mppm").

...

Ignored.

what

Character string indicating which quantity should be calculated. Either "vcov" for the variance-covariance matrix, "corr" for the correlation matrix, or "fisher" for the Fisher information matrix.

err

Character string indicating what action to take if an error occurs. Either "fatal", "warn" or "null".

Value

A numeric matrix (or NA or NULL).

Details

This is a method for the generic function vcov. The argument object should be a fitted multiple point process model (object of class "mppm") generated by mppm. The model must be a Poisson point process. The variance-covariance matrix of the parameter estimates is computed using asymptotic theory for maximum likelihood. If what="vcov" (the default), the variance-covariance matrix is returned. If what="corr", the variance-covariance matrix is normalised to yield a correlation matrix, and this is returned. If what="fisher", the Fisher information matrix is returned instead.

In all three cases, the rows and columns of the matrix correspond to the parameters (coefficients) in the same order as in coef{model}.

These calculations are not available if the model is not Poisson, or if it was computed using gam. In such cases, the argument err determines what will happen. If err="fatal" an error will occur. If err="warn" a warning will be issued and NA will be returned. If err="null", no warning is issued, but NULL is returned.

Examples

Run this code

data(waterstriders)
   fit <- mppm(Wat ~x, data=hyperframe(Wat=waterstriders))
   vcov(fit)