Modelling Frailty in the Manufacture of Respiratory Protective Devices.
Wassell JT; Kulczycki GW; Moyer ES
Proceedings of the Section on Epidemiology of the American Statistical Association 1994:133-138
Two frailty models, gamma and positive stable, were compared for testing organic vapor respirator cartridges to determine time until failure, or break through times. The 104 organic vapor respirator cartridges, filled with processed charcoal, used in the study were obtained from ten different manufacturers. To obtain unbiased estimates of covariate effects, methods were used to account for variability in unobserved differences in processes used by different manufacturers. Thirty of the cartridges were tested using carbon- tetrachloride, 36 were tested using pentane, and 38 were tested using hexane. Models were fitted for each challenge agent separately using the baseline Weibull regression and both gamma and positive stable frailty models. For the gamma frailty the likelihood ratio tests for the frailty parameters were all highly significant. For the positive stable frailty models, likelihood ratio tests for the frailty parameters were highly significant and even larger than the statistics for the fit of the gamma distributions. The software package Mathematica permitted both symbolic and numerical programming to be used to estimate parameters for very complicated likelihood functions. The authors conclude that frailty models provide a significantly better fit to the data than models (Weibull) assuming independence, with the positive stable model providing a better fit than the gamma frailty model.
Mathematical-models; Computer-models; Chemical-cartridge-respirators; Respiratory-protection; Respiratory-protective-equipment; Safety-equipment; Organic-vapors;
Proceedings of the Section on Epidemiology of the American Statistical Association