Comparison of the performances of quantile regression and parametric regression models in survival analysis

Comparison of the performances of Quantile Regression and Parametric Regression Models in Survival Analysis. In this study, it is aimed to compare the performances of QR and PR models in predicting survival time in survival dataset created by a simulation according to different sample sizes and the survival dataset which have homogeneous, heteregenous structure and is aimed to determine a model with minimum error. Besides, an algorithm is created by using R software for simulation practice in the application step of the study. By means of this algorithm; on real data sets and the data sets produced in heterogeneous structure, both PR models and QR models are compared in terms of the performance in survival analysis according to RMSE and MAE criteria. In the end of the simulation practice, it is determined that the errors of the QR model are minimum according to both two criteria and in different sample sizes. Among PR models, it is found that the model with minimum error according to RMSE is the lognormal regression; the model with minimum error according to MAE is log-logistic regression and the model with maximum error according to both two criteria is Weibull regression model. In the application of homogenous and heterogeneous breast cancer data sets; it is found that the performance of the QR method is better than PR models. For the homogeneous breast cancer data set and heterogeneous gbcs data set; according to RMSE and MAE criteria, the minimum error is found in QR model. Among PR models and according to RMSE and MAE criteria; in homogenous breast cancer data set, it is found that the model with minimum error is log-logistic; in heterogenous gbcs data set, it is found that the models with minimum error are log-normal and log-logistic, respectively. As a result; for both homogeneous and heteregeneous data sets, QR method has better performance than PR methods in survival analysis. Keywords: Parametric regression, Quantile regression, Simulation, Survival analysis.