Estimate Zero Inflated Poisson and Negative Binomial Models -----------------------------------------------------------
^zip^ depvar [indepvars] [^if^ exp] [^in^ range], [^lo^git^(^varlist^)^ ^l^evel(#) ^tr^ace ^nolog^ ^nobase^ ]
^zinb^ depvar [indepvars] [^if^ exp] [^in^ range], [^lo^git^(^varlist^)^ ^l^evel(#) ^tr^ace ^nolog^ ^nobase^ ]
^zippred^ varlist, [^xb^ | ^pr^ob ]
^zip^ and ^zinb^ share most features of estimation commands; see [U] 26 Estimation and post-estimation commands. However, to obtain predictions from the models, use ^zippred^ instead of @predict@.
Description -----------
^zip^ estimates Zero Inflated Poisson models to count data. ^zinb^ estimates Zero Inflated Negative Binomial models to count data. Both models allow for "excess zeros" in count models under the assumption that the population is characterized by two regimes, one where members always have zero counts, and one where members have zero or positive counts. The likelihood of being in either regime is estimated using a logit specification, while the counts in the second regime are estimated using a Poisson or Negative Binomial specification. See Long (1997) Chapter 8 and Greene (1997) for a discussion.
^zippred^ will generate predicted counts (using the default ^xb^ option) or the predicted probability of different counts (using ^pr^ob). ^zippred^ should be used instead of @predict@.
Options for ^zip^ and ^zinb^ ----------------------------
^logit(^varlist^)^ specifies the logit model for predicting membership in the zero-count regime. If varlist is not specified, the same model is used as for the count model.
^nolog^ supresses printing of the maximum likelihood iteration details.
^level(^#^)^ specifies the confidence level, in percent, for confidence intervals. The default is ^level(95)^ or as set by ^set level^.
^nobase^ suppresses estimation of the constant only baseline model. This is useful if convergence problems are encountered when estimating the baseline model.
Options for ^zippred^ ---------------------
^xb^ requests the predicted count. This is the default.
^pr^ob requests the predicted probabilities of different counts. The user should specify J+1 variables to obtain predicted counts for the integers from 0 to J. The first variable will contain the predicted probability, for each observation, of a zero count; the second will contain the probability of a count of one; and so on.
Examples --------
..zip derog income avgexp ownhome age, logit(income ownhome)
..zippred v0-v8, prob
References ----------
Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Sage.
Greene, William. 1997. Econometric Analysis. Prentice Hall.
Author ------
Jesper B. Sorensen University of Chicago jesper.sorensen@@gsb.uchicago.edu