Data with zeros are common
in a wide variety of fields including agriculture, biology, medicine,
public health, climatology, ecology, and more. Examples
include responses that are subject to a limit of detection (e.g.,
concentrations of pollutants in soil, water, or air); counts of the
number of plants or shoots or fruits in an agricultural experiment;
proportions of plants, animals, pine cones surviving to
maturity, affected by disease, etc.; number of insects of a given
species trapped or killed under different conditions; proportion of leaf
or fruit surface affected by disease; medical expenditures; rainfall
totals; hospital lengths of stay; motor vehicle fatalities;
and on and on . Such data present special challenges and are typically
not modeled adequately using traditional linear and generalized linear
models. Specialized models and techniques are often needed, such as
zero-inflated regression models, hurdle models,
two-part models, censored regression models, and other methods. In this
talk I will give examples of data with excess zeros and I will provide
an overview of statistical methodology for handling such data with an
emphasis on practical matters.
- Tags
-