Abstract:
|
A family of log-concave densities has many attractive features as it has some desirable statistical properties, contains many well-known statistical densities, and provides great flexibility. In this presentation, we propose a semi-parametric procedure to estimate parameters of an accelerated failure time model. To unearth the density of the error distribution, we applied the log-concave NPMLE for censored data. We develop the estimation algorithms and study the performance of the proposed methods to see the behavior of the parameter estimates compared to other existing methods especially for the cases in which the error distributions are skewed or have heavier tail than the normal.
|