|ISBN (digital version):||978-82-480-1168-2|
Accident count data are non-negative integers. A large part of the variation in such data is due to sheer and unexplainable randomness. There are strong reasons to believe that accident data are, at least approximately, Poisson distributed. Acknowledging this fact opens the door to an arsenal of quite efficient inference methods. We explain how these opportunities can be exploited – or missed.