Calculate Variance-Covariance Matrix for a Fitted Model...

vcov {stats}

R Documentation

Calculate Variance-Covariance Matrix for a Fitted Model Object

Description

Returns the variance-covariance matrix of the main parameters of a fitted model object. The “main” parameters of model correspond to those returned by coef, and typically do not contain a nuisance scale parameter (sigma).

Usage

vcov(object, ...)
## S3 method for class 'lm'
vcov(object, complete = TRUE, ...)
## and also for '[summary.]glm' and 'mlm'
## S3 method for class 'aov'
vcov(object, complete = FALSE, ...)

.vcov.aliased(aliased, vc, complete = TRUE)

Arguments

`object`	a fitted model object, typically. Sometimes also a `summary()` object of such a fitted model.
`complete`	for the `aov`, `lm`, `glm`, `mlm`, and where applicable `summary.lm` etc methods: logical indicating if the full variance-covariance matrix should be returned also in case of an over-determined system where some coefficients are undefined and `coef(.)` contains `NA`s correspondingly. When `complete = TRUE`, `vcov()` is compatible with `coef()` also in this singular case.
`...`	additional arguments for method functions. For the `glm` method this can be used to pass a `dispersion` parameter.

`aliased`	a `logical` vector typically identical to `is.na(coef(.))` indicating which coefficients are ‘aliased’.
`vc`	a variance-covariance matrix, typically “incomplete”, i.e., with no rows and columns for aliased coefficients.

Details

vcov() is a generic function and functions with names beginning in vcov. will be methods for this function. Classes with methods for this function include: lm, mlm, glm, nls, summary.lm, summary.glm, negbin, polr, rlm (in package MASS), multinom (in package nnet) gls, lme (in package nlme), coxph and survreg (in package survival).

(vcov() methods for summary objects allow more efficient and still encapsulated access when both summary(mod) and vcov(mod) are needed.)

.vcov.aliased() is an auxiliary function useful for vcov method implementations which have to deal with singular model fits encoded via NA coefficients: It augments a vcov–matrix vc by NA rows and columns where needed, i.e., when some entries of aliased are true and vc is of smaller dimension than length(aliased).

Value

A matrix of the estimated covariances between the parameter estimates in the linear or non-linear predictor of the model. This should have row and column names corresponding to the parameter names given by the coef method.

When some coefficients of the (linear) model are undetermined and hence NA because of linearly dependent terms (or an “over specified” model), also called “aliased”, see alias, then since R version 3.5.0, vcov() (iff complete = TRUE, i.e., by default for lm etc, but not for aov) contains corresponding rows and columns of NAs, wherever coef() has always contained such NAs.

[Package stats version 3.5.0 Index]