Coordinate descent methods for the penalized semiparametric additive hazards model

Anders Gorst-Rasmussen, Thomas Scheike

Research output: Book/ReportReportResearch

399 Downloads (Pure)

Abstract

For survival data with a large number of explanatory variables,
lasso penalized Cox regression is a popular regularization strategy. However,
a penalized Cox model may not always provide the best fit to data and can
be difficult to estimate in high dimension because of its intrinsic nonlinearity.
The semiparametric additive hazards model is a flexible alternative which is a
natural survival analogue of the standard linear regression model. Building on
this analogy, we develop a cyclic coordinate descent algorithm for fitting the
lasso and elastic net penalized additive hazards model. The algorithm requires
no nonlinear optimization steps and offers excellent performance and stability.
An implementation is available in the R-package ahaz and we demonstrate this
package in a small timing study and in an application to real data.
Original languageEnglish
PublisherDepartment of Mathematical Sciences, Aalborg University
Number of pages17
Publication statusPublished - 2011
SeriesResearch Report Series
NumberR-2011-10
ISSN1399-2503

Fingerprint

Additive Hazards Model
Coordinate Descent
Descent Method
Elastic Net
Penalized Regression
Cox Regression
Cox Model
Lasso
Descent Algorithm
Survival Data
Nonlinear Optimization
Linear Regression Model
Higher Dimensions
Analogy
Timing
Regularization
Nonlinearity
Analogue
Alternatives
Estimate

Keywords

    Cite this

    Gorst-Rasmussen, A., & Scheike, T. (2011). Coordinate descent methods for the penalized semiparametric additive hazards model. Department of Mathematical Sciences, Aalborg University. Research Report Series, No. R-2011-10
    Gorst-Rasmussen, Anders ; Scheike, Thomas. / Coordinate descent methods for the penalized semiparametric additive hazards model. Department of Mathematical Sciences, Aalborg University, 2011. 17 p. (Research Report Series; No. R-2011-10).
    @book{03c1c4a6d8c44ad3aab3ccccfe82c236,
    title = "Coordinate descent methods for the penalized semiparametric additive hazards model",
    abstract = "For survival data with a large number of explanatory variables,lasso penalized Cox regression is a popular regularization strategy. However,a penalized Cox model may not always provide the best fit to data and canbe difficult to estimate in high dimension because of its intrinsic nonlinearity.The semiparametric additive hazards model is a flexible alternative which is anatural survival analogue of the standard linear regression model. Building onthis analogy, we develop a cyclic coordinate descent algorithm for fitting thelasso and elastic net penalized additive hazards model. The algorithm requiresno nonlinear optimization steps and offers excellent performance and stability.An implementation is available in the R-package ahaz and we demonstrate thispackage in a small timing study and in an application to real data.",
    keywords = "survival, additive hazards, lasso, elasti net, coordinate descent",
    author = "Anders Gorst-Rasmussen and Thomas Scheike",
    year = "2011",
    language = "English",
    publisher = "Department of Mathematical Sciences, Aalborg University",

    }

    Gorst-Rasmussen, A & Scheike, T 2011, Coordinate descent methods for the penalized semiparametric additive hazards model. Research Report Series, no. R-2011-10, Department of Mathematical Sciences, Aalborg University.

    Coordinate descent methods for the penalized semiparametric additive hazards model. / Gorst-Rasmussen, Anders; Scheike, Thomas.

    Department of Mathematical Sciences, Aalborg University, 2011. 17 p.

    Research output: Book/ReportReportResearch

    TY - RPRT

    T1 - Coordinate descent methods for the penalized semiparametric additive hazards model

    AU - Gorst-Rasmussen, Anders

    AU - Scheike, Thomas

    PY - 2011

    Y1 - 2011

    N2 - For survival data with a large number of explanatory variables,lasso penalized Cox regression is a popular regularization strategy. However,a penalized Cox model may not always provide the best fit to data and canbe difficult to estimate in high dimension because of its intrinsic nonlinearity.The semiparametric additive hazards model is a flexible alternative which is anatural survival analogue of the standard linear regression model. Building onthis analogy, we develop a cyclic coordinate descent algorithm for fitting thelasso and elastic net penalized additive hazards model. The algorithm requiresno nonlinear optimization steps and offers excellent performance and stability.An implementation is available in the R-package ahaz and we demonstrate thispackage in a small timing study and in an application to real data.

    AB - For survival data with a large number of explanatory variables,lasso penalized Cox regression is a popular regularization strategy. However,a penalized Cox model may not always provide the best fit to data and canbe difficult to estimate in high dimension because of its intrinsic nonlinearity.The semiparametric additive hazards model is a flexible alternative which is anatural survival analogue of the standard linear regression model. Building onthis analogy, we develop a cyclic coordinate descent algorithm for fitting thelasso and elastic net penalized additive hazards model. The algorithm requiresno nonlinear optimization steps and offers excellent performance and stability.An implementation is available in the R-package ahaz and we demonstrate thispackage in a small timing study and in an application to real data.

    KW - survival

    KW - additive hazards

    KW - lasso

    KW - elasti net

    KW - coordinate descent

    M3 - Report

    BT - Coordinate descent methods for the penalized semiparametric additive hazards model

    PB - Department of Mathematical Sciences, Aalborg University

    ER -

    Gorst-Rasmussen A, Scheike T. Coordinate descent methods for the penalized semiparametric additive hazards model. Department of Mathematical Sciences, Aalborg University, 2011. 17 p. (Research Report Series; No. R-2011-10).