Testing for the Presence of a Zero Hurdle

Description

Wald test of the null hypothesis that no zero hurdle is required in hurdle regression models for count data.

Usage

hurdletest(object, ...)

Arguments

object	A fitted model object of class “hurdle” as returned by hurdle, see details for more information.
…	arguments passed to linearHypothesis.

Details

If the same count distribution and the same set of regressors is used in the hurdle model for both, the count component and the zero hurdle component, then a test of pairwise equality between all coefficients from the two components assesses the null hypothesis that no hurdle is needed in the model.

The function hurdletest is a simple convenience interface to the function linearHypothesis from the car packages that can be employed to carry out a Wald test for this hypothesis.

Value

An object of class “anova” as returned by linearHypothesis.

References

Cameron AC, Trivedi PK (1998). Regression Analysis of Count Data. New York: Cambridge University Press.

Cameron AC, Trivedi PK (2005). Microeconometrics: Methods and Applications. Cambridge: Cambridge University Press.

See Also

hurdle, linearHypothesis

Examples

library("countreg")

data("CrabSatellites", package = "countreg")
cs <- CrabSatellites[, c("satellites", "width", "color")]
cs$color <- as.numeric(cs$color)
fm <- hurdle(satellites ~ ., data = cs, dist = "negbin", zero = "negbin")
if(require("car")) hurdletest(fm)

Wald test for hurdle models

Restrictions:
count_((Intercept) - zero_(Intercept) = 0
count_width - zero_width = 0
count_color - zero_color = 0

Model 1: restricted model
Model 2: satellites ~ .

  Res.Df Df  Chisq Pr(>Chisq)    
1    168                         
2    165  3 50.362  6.689e-11 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1