Statistical methods in risk evaluation

This paper is concerned with the statistical methods involved in the evaluation of fire risk in an industry.The risk could be expressed in terms of two main components—the probability of fire starting and the probability of damage reaching various levels in a fire. The latter probability involves the concept of probability distribution of fire loss, the nature of which is discussed with reference to statistical and physical considerations. The values of the parameters of this distribution and the probability of fire starting will vary depending on factors such as sources of ignition, size of the property, and the presence or absence of sprinklers.It is shown how results for a group of similar properties could be modified for evaluating the risk in an individual property. For this purpose, detailed particulars of the property and processes will be required. If individual figures for financial losses are available only for large fires the parameters of the probability distribution of fire loss can be estimated by applying the statistical theory of extreme values. Basic principles of this theory and its practical uses are explained. As a project for further research it is proposed that a model based on stochastic processes should be developed for predicting the spread of fire in a building. An outline of such a model is given specifying the data required for estimating and validating the parameters of the model.