The road accident literature provides no universally accepted
definition of a road accident ‘hotspot’. Hauer (1997) describes how
researchers rank locations according to accident rate while other
researchers use accident frequencies (accident per road kilometre).
Road accident hotspot analysis has traditionally centred on
road segments or specific junctions (Thomas, 1996)while area wide
hotspots and the spread of risk which is produced from a collision
is somewhat neglected. Traditionally since the late 1970s statistical
models have been applied to road accident analysis, however the
early models where flawed in such that they would assume accidents
to be normally distributed (Oppe, 1979; Ceder and Livneh,
1982). The next stage was to accommodate this statistical drawback
was to use Poisson log linear regression to account for the
randomness of accidents in time and space (Blower et al., 1993).
Many authors such as Hauer and Persaud (1987), Miaou (1994),