Parameter estimation of mean survival time using parametric and nonparametric approaches

Abstract

Exploring health related quality of life is usually the focus of survival studies. Using the data of breast cancer, an investigation about the mean survival time of cancer patients was explored, using the nonparametric and parametric modeling approaches. The Kaplan-Meier method and three of the distribution were considered in this study which is Weibull distribution, exponential distribution and lognormal distribution. Other than that, the Anderson Darling test is used to test if a sample data came from a population with a specific distribution. Based on the result, the data came from a Weibull distribution because the distribution has the minimum Anderson-Darling (adjusted) value. The simulation study has been done to see the efficiency of parametric and nonparametric estimator by observing the Relative Efficiency (RE) values. The results show that parametric estimator provide better estimates than the Kaplan-Meier estimator if the correct distribution is assumed.