1
$\begingroup$

How can I calculate the deviance for the factors and the null and saturated models with Hurdle models?

I used the function hurdle() from the package pscl. The functions anova() or deviance() are not applicable.

$\endgroup$
4
$\begingroup$

As the hurdle and zero-inflation models are both composed of two submodels they do not have all the properties of classical GLMs (generalized linear models) like the Poisson model. That's why we decided not to provide a deviance() or anova() method for "hurdle" and "zeroinfl" objects. Instead there is a logLik() method for extracting the log-likelihood (which is closely connected to the deviance in GLMs). For inference, functions like lrtest() from lmtest for the likelihood ratio test can be used. Further options are discussed in vignette("countreg", package = "pscl") (see e.g. the end of Section 2.2).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.