A Note on the Stochastic EM Algorithm Based on Left Truncated Right Censored Data From Burr XII Distribution

Abstract

The Burr XII distribution is a flexible model for failure-time data. A very general and commonly observed structure for failure-time data involves left truncation and right censoring. In this article, modelling of left truncated right censored failure-time data by the Burr XII distribution is discussed. The steps of the stochastic expectation maximization algorithm, which is a useful technique of estimation for incomplete data structures, are developed to estimate the model parameters of the Burr XII distribution. The Newton-Raphson method, which is a direct method of obtaining maximum likelihood estimates by optimizing the observed likelihood is also used. The two methods of inference are assessed and compared through a Monte Carlo simulation study. Discussions of the inferential methods are extended to the cases of a three-parameter Burr XII model, and a covariate-included model. An illustrative example based on real data is provided. Finally, an application of the inferential results to a prediction issue is discussed with an illustration.