In this article we propose a parsimonious parameterisation to model the so-called erosion of the covariate effect in the Cox model, namely a covariate effect approaching to zero as the follow-up time increases. The proposed parameterisation is based on the segmented relationship where proper constraints are set to accomodate for the erosion. Relevant hypothesis testing is discussed. The approach is illustrated on two historical datasets in the survival analysis literature, and some simulation studies are presented to show how the proposed framework leads to a test for a global effect with good power as compared with alternative procedures. Finally, possible generalisations are also presented for future research.

Muggeo, V., & Attanasio, M. (2011). Segmented relationships to model erosion of regression effect in Cox regression. STATISTICAL METHODS IN MEDICAL RESEARCH, 20(4), 401-414 [10.1177/0962280209358643].

Segmented relationships to model erosion of regression effect in Cox regression

MUGGEO, Vito Michele Rosario;ATTANASIO, Massimo
2011

Abstract

In this article we propose a parsimonious parameterisation to model the so-called erosion of the covariate effect in the Cox model, namely a covariate effect approaching to zero as the follow-up time increases. The proposed parameterisation is based on the segmented relationship where proper constraints are set to accomodate for the erosion. Relevant hypothesis testing is discussed. The approach is illustrated on two historical datasets in the survival analysis literature, and some simulation studies are presented to show how the proposed framework leads to a test for a global effect with good power as compared with alternative procedures. Finally, possible generalisations are also presented for future research.
Settore SECS-S/01 - Statistica
Settore SECS-S/05 - Statistica Sociale
Muggeo, V., & Attanasio, M. (2011). Segmented relationships to model erosion of regression effect in Cox regression. STATISTICAL METHODS IN MEDICAL RESEARCH, 20(4), 401-414 [10.1177/0962280209358643].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/10447/44561
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 1
social impact