This study sets out to investigate the potential of sequence analysis in road accident case studies, including some more sophisticated methods than were available to Malaterre (who wanted to see the approach further refined). Police accident records were chosen as the data source -- although they were not expected to provide a complete record (see Grayson and Hakkert, 1987, for comments on the reliability and validity of police accident data). It was hoped that if a methodology could be produced which worked effectively on existing police data then a colossal new data base would become available for accident research of this kind. Two hundred police accident reports were randomly cted from Nottingham Constabulary's records for 1988 (consisting of 100 right turns off a main road and 100 right turns onto a main road). Both static features (such as weather and road conditions, time of day, carriageway type, and so on) and the sequence of events were coded. The static features were unproblematic to code, but it was necessary to devise a special-purpose coding scheme which could cope with the sequential information, using a traffic related action analysis language (TRAAL) to encode the actions of the right-turner. So far, most of the effort in this project has gone into collecting, transcribing, encoding and checking data; but some preliminary results have already been obtained using three main techniques -- rule-finding, sequential analysis and conventional statistics. The remainder of this paper examines various possible approaches to road accident research, including generalization, aggregation, cumulative case study, behavioural grammars, syndromes etc.
Abstract