Yaşam çözümlemesine yeni bir yaklaşım: MARS

No Thumbnail Available

Date

2011

Journal Title

Journal ISSN

Volume Title

Publisher

Sağlık Bilimleri Enstitüsü

Abstract

In this dissertation study it was investigated that whether Multivariate adaptive regression splines (MARS) can be used in survival analysis or not. In this context, MARS method was introduced briefly and its application to survival analysis was shown with its mathematical infrastructure.In order to adapt the MARS method to survival analysis, Martingale and Deviance residuals of Cox model does not use independent variable was investigated. Martingale and Deviance residuals were assumed as dependent variables of MARS method, so that it was tried to construct a model by using an independent variable which represents hazard and combines the censored variables.To show the applicability of MARS model, cancer data were used. In this carcinoma data set the variables having risk effect on variables was determined and the results obtained for different cut points for Platelet were used to improve the success of classical Cox-PH model.The method introduced in this thesis study was applied on UMARU-UIS data set which is frequently come across with the references related to survival analysis and it was shown that MARS model can provide more contributions to classical Cox-PH model.Key Words: Cox-PH model, Deviance residuals, MARS, Martingale residuals, Multivariate adaptive regression splines

Description

Keywords

MARS, Yaşam çözümlemesi

Citation