View Item 
      •   IPB Repository
      • Dissertations and Theses
      • Undergraduate Theses
      • UT - School of Data Science, Mathematic and Informatics
      • UT - Statistics and Data Sciences
      • View Item
      •   IPB Repository
      • Dissertations and Theses
      • Undergraduate Theses
      • UT - School of Data Science, Mathematic and Informatics
      • UT - Statistics and Data Sciences
      • View Item
      JavaScript is disabled for your browser. Some features of this site may not work without it.

      Evaluation of Cox Regression Performance on Violation of Proportional Hazards Assumption Using Repeated Sampling on Breast Cancer Cases

      Thumbnail
      View/Open
      Cover (371.8Kb)
      Fulltext (1.804Mb)
      Lampiran (1.766Mb)
      Date
      2025
      Author
      Kireinahana, Kaylila
      Indahwati
      Afendi, Farit Mochamad
      Metadata
      Show full item record
      Abstract
      Survival analysis is crucial in medical research to model the time until an event occurs, such as death or recovery. The Cox Proportional Hazards (Cox PH) model is widely used; however, it depends on the proportional hazards (PH) assumption, which is often violated in practice, especially when the effect of a covariate changes over time. The stratified Cox model and the extended Cox model are two approaches frequently used to address violations of the PH assumption. In this study, the performance of the two models were compared. This study also evaluated the effect of sample size variations on model robustness and the ability to detect violations of the PH assumption. Model stability and the PH assumption test using Schoenfeld residuals were evaluated through repeated stratified sampling. Based on data from 4024 breast cancer cases in women collected by the National Cancer Institute (NCI) from 2006–2010, violations of the assumption were found in two binary variables, such as estrogen and progesterone receptor status. The extended Cox model with the Heaviside time function performed best, with a C-index of 0,741 and an AIC of 9373,679. Repeated sampling results showed that small sample sizes produced unnaturally high C-index values and masked assumption violations. In contrast, larger sample sizes provided more stable estimates and violation detection. These results highlighted the importance of testing model assumptions and selecting an appropriate survival model, particularly for data sets with limited size and unbalanced proportions.
       
      Analisis daya tahan (survival analysis) krusial dalam penelitian medis untuk memodelkan waktu hingga suatu peristiwa terjadi, seperti kematian atau kesembuhan. Model Cox Proportional Hazards (Cox PH) merupakan model yang sering digunakan, namun model tersebut bergantung pada asumsi proportional hazards (PH) yang dalam praktiknya kerap dilanggar, terutama ketika efek kovariat berubah terhadap waktu. Model Cox stratifikasi dan model extended Cox merupakan dua pendekatan yang sering digunakan untuk mengatasi pelanggaran asumsi PH. Dalam peneltian ini performa kedua model akan dibandingkan. Studi ini juga mengevaluasi pengaruh variasi ukuran sampel terhadap ketahanan model dan kemampuan dalam mendeteksi pelanggaran asumsi PH. Kestabilan model dan uji asumsi PH melalui sisaan Schoenfeld dievaluasi melalui stratified sampling berulang. Berdasarkan 4024 data kasus kanker payudara pada perempuan yang dikumpulkan oleh National Cancer Institute (NCI) periode 2006–2010, ditemukan pelanggaran asumsi pada dua peubah biner, yaitu status reseptor estrogen dan progesteron. Model extended Cox dengan fungsi waktu Heaviside menunjukkan hasil terbaik, dengan nilai C-index 0,741 dan AIC 9373,679. Hasil pengambilan sampel berulang menunjukkan bahwa ukuran sampel yang kecil cenderung menghasilkan nilai C-index yang lebih tinggi secara tidak wajar dan menyamarkan pelanggaran asumsi, sementara ukuran sampel yang lebih besar memberikan estimasi yang lebih stabil dan kemampuan deteksi atas pelanggaran. Hasil ini menekankan pentingnya pengujian asumsi model dan pemilihan model survival yang tepat, khususnya pada gugus data dengan ukuran terbatas dan proporsi yang tidak seimbang.
       
      URI
      http://repository.ipb.ac.id/handle/123456789/169702
      Collections
      • UT - Statistics and Data Sciences [82]

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository
        

       

      Browse

      All of IPB RepositoryCollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

      My Account

      Login

      Application

      google store

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository