sandwich.vcov and score.vcov are two similar estimators
for the variance-covariance matrix of a probabilistic index model.
These functions are meant to be used within a call to pim
as a value for the argument vcov.
sandwich.vcov(fitted, X, Y, W, link, poset, ...)score.vcov(fitted, X, Y, W, link, poset, ...)
The fitted values (calculated as X %*% coef with
X the design matrix and coef the coefficients)
the design matrix
a numeric vector with pseudoresponses
a numeric vector with weights. If weights are not applicable,
set to NULL (the default)
a character vector with the link function
a list with the left and right indices. See poset
for more information.
arguments passed to downstream methods.
the variance-covariance matrix
You can create your own estimating functions for the variance-covariance
matrix. To do so, you have to make sure that your function allows for
the exact same arguments. As the function pim.fit calculates the
fitted values already, there's no need to incorporate the calculation of
these inside the function.
sandwich.estimator for more information on the
actual fitting process. pim for a few examples in how
these are used