A logistic reanalysis of Ashford and Sowden's data on respiratory symptoms in British coal miners.

A model based on Mantel's [1966] generalization of the logistic model to polychotomous dosage-response curves is presented and applied to data on respiratory symptoms in British coal miners. The model allows for the inclusion of an associated occurrence between the two respiratory symptoms. Various models with differing parametric assumptions are fitted to the data by maximum likelihood and are compared with one another by chi-square statistics. The use of logistic analysis for the intercomparison of several sample normal means and variances and its extension to the case of censoring is discussed.