ANALISIS PERBANDINGAN MODEL COX PROPORTIONAL HAZARDS DAN AALEN’S ADDITIVE PADA KUMPULAN DATA KANKER PROSTAT BERBASIS WEB APPLICATION

Oleh : Alexander Indrajaya Limantoro, Diana Syarifah Permai dan Muhammad Fitra Kacamarga

The goals of this research are to search the significance of covariates of prostate cancer data set by Cox proportional hazards and Aalen’s additive models, and to determine which model fits better. The data related to 481 male patients, diagnosed with prostate cancer stage 3 & 4 with ten covariates significances to be checked. Cox PH, and Aalen’s Additive are used to test the significance of covariates, and 10 – fold cross validation with c – index is used to determine which model is better. Four covariates are not significant to the event by both models, and the 10 – fold cross validation scores of both models do not differ much. Six covariates have proved significant to the event by the Cox Model, they are ‘physical fitness’, ‘recipe’, ‘age’, ‘combined index of stage and histologic grade’, ‘serum hemoglobin’, ‘size of primary tumor’. Similar result is given by Aalen’s additive model. Based on 10 – fold cross validation with c – index, Aalen’s additive is a better model than Cox PH on prostate cancer data set. The Aalen’s additive fits better than Cox PH on prostate cancer data set, but in general, both Cox PH and Aalen’s additive give different piece of information. (AL)

Keywords : Aalen’s additive model, cox proportional hazards model , prostate cancer, survival analysis, survival regression