| Æ¡Æ¡In this talk, we extend the results from Jensen (1993) and prove that the distribution of signed root log-likelihood ratio statistic can be approximated by its bootstrap distribution up to second order accuracy when data is censored. We also use a simulation study in constructing confidence intervals to investigate the adequacy of the approximation provided by the theoretical result. The bootstrap-t and $BC_{a}$ methods are second order accurate when the data are complete. Our simulation results show that the method based on bootstrap signed root log-likelihood ratio statistic outperforms the bootstrap-t and $BC_{a}$ methods in constructing one-sided confidence bounds when the data are Type I censored. |