Answers are grounded in this paper and cite their sources.
A Bates-Neyman model is applied to four sets of accident data obtained in a study of childhood accidents. In each case, the model fits the data well. The inference is therefore tenable that the assumptions on which the model is based describe the processes underlying the observed accident distribution. For our purpose, the crucial assumptions are, first, that there are differences in accident liability within each of the groups of children studied and, second, that these differences persist over time. The model is then used to estimate the extent to which a measure of accident liability based on observed number of accidents would be contaminated by chance factors. We find as expected, that some children are likely to be classified incorrectly with respect to liability. However, the estimated proportion of incorrect classifications is low enough (especially in the extreme groups) to warrant using number of accidents as a criterion of accident liability. It is also demonstrated that keeping contamination to acceptably low levels requires rigorous accident criteria, large groups from which to select accident repeater study groups, and accident histories for prospective subjects extending over long periods of time.
No takes yet. Share an insight, caveat, or question.
Mellinger et al. (1965) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: