Calculate Variance-Covariance Matrix for a `nlregb' Fit
Usage
vcov.nlregb(object, method=c("Fisher", "observed", "Huber"),
scale=object$scale, eps=0.001, tol=1)
Arguments
object
|
The return values of a nlregb fit.
|
method
|
The theoretical basis for the estimate. This can be based on the
Fisher information (the usual assumption) or the observed information
assuming the model is true or a Huber-White sandwich estimator
which allows the model to be false. Only the Fisher method is
available if the model was fitted without gradient (Jacobian) information.
|
scale
|
An initial scaling for the parameters.
|
eps
|
The step size (as a multiple of min(1, abs(param)) ) for
finite-difference approximations to terms in the Hessian.
|
tol
|
Relative change in sum-of-squares sought in a local quadratic
approximation. See the code for the scaling used.
|
Description
Method for vcov
to find the variance-covariance matrix of the
coefficients of fits by nlregb
.Value
A matrix of the estimated covariances between the parameter estimates
in the non-linear regression.See Also
vcov