I am working with a dataset df which comprises count data count and a number of categorical variables. df looks something like this:

count    var1    var2    var3    var4    var5    var6  
12       blue    X       old     UK      yes     no
7        red     Y       old     USA     no      no 
14       green   X       new     UK      no      no

To assess the distribution of count values, I am using a hurdle model from the pscl package:

model = hurdle(count ~ var1 + var2 + var3 + var4 + var5 + var6,  data=df, link="logit", dist="negbin")

In the interests of optimal model selection (i.e. determining which combination of variables best fits the data), I would like to compute a measure of explained variance for each variable in the above model. For a simple linear regression I would compute R-squared using the lm() method. Is there some equivalent measure/method that can be used for hurdle models?


1 Answer 1


Hurdle models consist of two parts: a binomial model for modeling the probability that a zero is observed and, in your case, a truncated negative binomial model for modeling the non-zero observations. Each part can have covariates and the covariates don't have to be the same in each part. The approach used by Zuur et al. (2009) for selection of explanatory variables is to drop each term in turn and select the optimal model using the likelihood ratio statistic or AIC.

Zuur, A.F., E.N. Ieno, N.J. Walker, A.A. Saveliev, and G.M. Smith. 2009. Mixed effects models and extensions in ecology with R. Springer-Verlag, New York. 574 p.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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