Inference in the log-linear Birnbaum-Saunders Model under type I Censoring

摘要

In thispaper inference for a log-linear Birnbaum-Saunders model under Type I censoringis presented. Methods of inference based on maximum likelihood, includingnormal approximation, profile likelihood, signed deviance statistics, as wellas parametric bootstrap are presented. Inference for both shape andregression parameters are studied, as well as quantiles and survivalprobabilities. Results of a simulation study to compare small sample accuracyof the various approaches are discussed and two examples with real data areshown.