| Summary: | 碩士 === 國立臺北大學 === 統計學系 === 106 === In this study, we use Paik & Ying's (2012) two-stage maximum likelihood estimation method to obtain the correlation parameters between the same clusters through individual messages. Therefore, we want to estimate parameters in the same way under the assumption of two different copulas. In this paper, the survival time is simulated using the Clayton copula and the Frank copula in the Archimedes copula. In the data set, there are 2 to 4 surviving events within the same cluster, to explore the intercept parameters, regression coefficients, shape parameters and the dependence on survival time under different copulas. We compare whether the estimation error will be different within the same cluster because of different number of observations or different proportion of censored data. Finally, the standard deviation and standard error of the estimated values of various parameters were observed. The bootstrap method was used to take repeated sampling from the generated data, we construct 95% confidence intervals and calculate the coverage probability. According to the results, the two-stage maximum likelihood estimation method proposed is feasible.
|
|---|
