Considering the cyclical nature of accident and injury data, it is reasonable to consider the use of time- series analysis for modeling these data. One approach involved fitting a Box- Jenkins, auto- regressive, moving- average model to the data and using the model to forecast future values. A second approach utilized sine or cosine models to fit the cyclical pattern. A comparison of the two models, for a set of injury data in a glass manufacturing facility, indicated a clear superiority of the Box- Jenkins approach, not only for fitting a seasonal cycle, but also for accommodating monthly trends.
Abstract