Variance-covariance matrix — vcov.coxph • survival

Extract and return the variance-covariance matrix.

Usage

# S3 method for class 'coxph'
vcov(object, complete=TRUE, ...)
# S3 method for class 'coxphms'
vcov(object, complete=TRUE, matrix=FALSE, ...)
# S3 method for class 'survreg'
vcov(object, complete=TRUE, ...)

Arguments

object: a fitted model object
complete: logical indicating if the full variance-covariance matrix should be returned. This has an effect only for an over-determined fit where some of the coefficients are undefined, and coef(object) contains corresponding NA values. If complete=TRUE the returned matrix will have row/column for each coefficient, if FALSE it will contain rows/columns corresponding to the non-missing coefficients. The coef() function has a similar complete argument.
matrix: if TRUE the result will be an array with one covariance matrix per transition in the model; vcov(object, matrix=TRUE)[,,i] will be the variance matrix corresponding to coef(object, matrix=TRUE)[,i]
...: additional arguments for method functions

Value

a matrix or array

Details

For the coxph and survreg functions the returned matrix is a particular generalized inverse: the row and column corresponding to any NA coefficients will be zero. This is a side effect of the generalized cholesky decomposion used in the unerlying compuatation.