LIKELIHOOD RATIO-BASED CONFIDENCE-INTERVALS IN SURVIVAL ANALYSIS

Confidence intervals for the survival function and the cumulative haza rd function are considered. These confidence intervals are based on an inversion of the likelihood ratio statistic. To do this, two extensio ns of the likelihood, each of which yields meaningful likelihood ratio hypothesis tests and subsequent confidence intervals, are considered. The choice of the best extension is difficult. In the failure time se tting, the binomial extension is best in constructing confidence inter vals concerning the survival function and the Poisson extension is bes t in constructing confidence intervals concerning the cumulative hazar d. Simulations indicate that these two methods perform as well as or b etter than competitors based on the asymptotic normality of the estima tor.