pscl (version 0.59)

zeroinfl: zero-inflated regression models for count data

Description

fit zero-inflated regression models for count data via maximum likelihood

Usage

zeroinfl(count = y ~ ., x = ~1, z = ~1,
        data = list(),
        link = "logit", dist = "poisson",
        method = "BFGS", trace = FALSE, maxit = 50000,
        na.action = na.omit)

Arguments

count

count being modeled, passed as the left-hand side of a formula; lowest count must be zero for zero-inflated models

x

right-hand side of a formula giving covariates for the count part of the model

z

right-hand side of a formula giving covariates for the zero-inflated part of the model

data

a data frame

link

link function for zero-inflated part of the model (choices are logit (default) or probit)

dist

type of count model, "poisson" (default) or "negbin"

method

method for maximizing the log-likelihood function, only "BFGS" and "Nelder-Mead" are supported

trace

logical, if TRUE, display progress of maximization

maxit

maximum number of iterations in maximization

na.action

method for handling missing data, default is na.omit

Value

• an object of class zeroinfl, i.e., a list with components including
• stval
  start values used in optimzation
• par
  Maximum likelihood estimates
• hessian
  Matrix of second derivatives of the log-likelihood function evaluated at the MLEs; computed numerically by optim
• llh
  value of the log-likelihood function at the MLEs
• y
  vector of counts actually fitted (after any screeing of missing data)
• x
  matrix of covariates used in fitting the count model
• z
  matrix of covariates used in fitting the zero-inflated component

Details

Zero-inflated count models are a type of two-component mixture model, with a component for zero counts, and the other component for the positive counts. Poisson or negative-binomial models are used for the count component of the model; logit or probit is typically used to model the probability of a zero count. Covariates are used to predict each outcome. optim is used to find maximum likelihood estimates and to compute a Hessian matrix after convergence.

References

Lambert, Diane. 1992. "Zero-Inflated Poisson Regression, With an Application to Defects in Manufacturing." Technometrics.V34(1):1-14

Cameron, A. Colin and Pravin K. Trevedi. 1998. Regression analysis of count data. New York: Cambridge University Press.

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Number 7 in Advanced Quantitative Techniques in the Social Sciences. Thousand Oaks, California: Sage.

See Also

summary.zeroinfl, predict.zeroinfl, hurdle, glm.nb

Methods are supplied for the generic functions coef and logLik, for objects of class "zeroinfl".

Examples

data(bioChemists)
zip <- zeroinfl(count=art ~ .,
                x = ~ fem + mar + kid5 + phd + ment,   ## no z model
                data=bioChemists,trace=TRUE)
zip
summary(zip)
zip2 <- zeroinfl(count=art ~ .,
                x = ~ fem + mar + kid5 + phd + ment,
                z = ~ fem + mar + kid5 + phd + ment,
                data=bioChemists,trace=TRUE)
zip2
summary(zip2)

zinb <- zeroinfl(count=art ~ .,
                 x = ~ fem + mar + kid5 + phd + ment,
                 z = ~ fem + mar + kid5 + phd + ment,
                 dist="negbin",
                 data=bioChemists,trace=TRUE)
zinb
summary(zinb)