Modeling Nonnegative Data with Clumping at Zero A Survey.pdf

JIRSS (2002) Vol. 1, Nos. 1-2, pp 7-33 Modeling Nonnegative Data with Clumping at Zero: A Survey Yongyi Min, Alan Agresti Department of Statistics, University of Florida, Gainesville, Florida, USA 32611-8545. (ymin@stat.ufl.edu, aa@stat.ufl.edu) Abstract. Applications in which data take nonnegative values but have a substantial proportion of values at zero occur in many dis- ciplines. The modeling of such “clumped-at-zero” or “zero-inflated” data is challenging. We survey models that have been proposed. We consider cases in which the response for the non-zero observations is continuous and in which it is discrete. For the continuous and then the discrete case, we review models for analyzing cross-sectional data. We then summarize extensions for repeated measurement analyses (e.g., in longitudinal studies), for which the literature is still sparse. We also mention applications in which more than one clump can oc- cur and we suggest problems for future research. 1 Introduction In some applications, the response variable can take any nonnegative value but has positive probability of a zero outcome. We refer to a variable as semicontinuous when it has a continuous distribution Received: May 2002 Key words and phrases: Compliance, finite mixture model, logistic regression, Neyman type a distribution, proportional odds model, semicontinuous data, Tobit model, zero-inflated data. 8 Min and Agresti except for a probability mass at 0. Semicontinuous data are common in many areas. For example, when each observation is a record of the total rainfall in the previous day, many days have no rainfall. In a study of household expenditures, some households spend nothing on a certain commodity during the period of investigation. In a study of annual medical costs, a portion of the