In this paper inference for a log-linear Birnbaum-Saunders model under Type I censoring is presented. Methods of inference based on maximum likelihood, including normal approximation, profile likelihood, signed deviance statistics, as well as parametric bootstrap are presented. Inference for both shape and regression parameters are studied, as well as quantiles and survival probabilities. Results of a simulation study to compare small sample accuracy of the various approaches are discussed and two examples with real data are shown.
ISSN: 2241-0376 (Online)