The majority of RTM studies to‐date have tested risk terrain models through logistic regression with shooting
incidents as the dependent variable (Caplan et al., 2011; Kennedy et al., 2011). In logistic regression, the
dependent variable is dichotomized to represent either the presence (“1”) or absence (“0”) of a particular feature.
In the case of shootings, logistic regression tests the influence of the independent variable(s) (e.g. “risk values”) on
the presence or absence of any shooting incidents. Given the infrequent occurrence of shootings (compared to
other crime types), and the fact that most spatial units are unlikely to have more than 1 incident, logistic
regression is an appropriate statistical test in such cases. However, for more frequently‐occurring crime types,
logistic regression may undercount the total number of crimes since multiple incidents are collapsed into a single
unit to fulfill the requirements of logistic regression. Such undercounting of incidents may depreciate the validity
of the model, particularly by underestimating the predictive capacity of risk terrain models. This brief discusses the
use of count regression models, namely Poisson and negative binomial regression, as a method of overcoming the
limitations of logistic regression in RTM studies focusing on more frequently‐occurring crime types.