Bread matrix for sandwich estimator

Returns the bread component B such that the model-based variance is B/n. Specifically: \(B = n (Z' W Z)^{-1}\) where Z is the model matrix and \(W = \mathrm{diag}(\mathrm{bread\_weights})\).

Usage

# S3 method for class 'uncounted'
bread(x, ...)

Arguments

x

An uncounted object

...

Ignored

Note

For NB models, these methods operate on the mean-model parameters (alpha, beta, and optionally gamma) only, excluding theta. The stored vcov() on NB objects uses a dedicated theta-aware path instead. Calling sandwich::vcovHC() directly on an NB object gives theta-conditional standard errors, which differ from vcov().