Frailty models of manufacturing effects.
Wassell-JT; Kulczycki-GW; Moyer-ES
Lifetime Data Anal 1995 Jun; 1(2):161-170
Gamma and positive frailty models of the distribution of breakthrough times of 104 organic vapor respirator cartridges produced by ten different manufacturers were compared. Test conditions compared breakthrough times of carbon-tetrachloride (56235) (CCl4) to those of pentane (109660) (C5H12) and n-hexane (110543) (C6H14) as challenge agents. A review of frailty models was provided. Baseline Weibull regression and gamma and positive stable frailty models were fit separately for each challenge agent. Likelihood ratio tests indicated that significantly better fits were calculated with both frailty models than with Weibull regression. The positive frailty models provided the best fit to the data with breakthrough times of 95.71, 75.98 and 70.14 minutes for CCl4, C5H12, and C6H14, respectively. Although cartridge sorbent weight was significant in the Weibull and fixed effects models, this factor was not significant in the frailty models due to variations in sorbent capacity between manufacturers. Fixed effects models using Cox and Weibull regression were used to compare covariants for different manufacturers. The authors conclude that the positive stable frailty model provides a better fit to the breakthrough data.
Mathematical-models; Personal-protective-equipment; Respiratory-protective-equipment; Materials-testing; Organic-vapors; Aliphatic-hydrocarbons; Failure-analysis; Chlorinated-hydrocarbons;
Author Keywords: Survival analysis; Cox regression; random effects; Weibull distribution
56-23-5; 109-66-0; 110-54-3
Lifetime Data Analysis