Local linear estimation of concordance probability with application to covariate effects models on association for bivariate failure-time data

Aidong Adam Ding*, Jin Jian Hsieh, Weijing Wang

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

Bivariate survival analysis has wide applications. In the presence of covariates, most literature focuses on studying their effects on the marginal distributions. However covariates can also affect the association between the two variables. In this article we consider the latter issue by proposing a nonstandard local linear estimator for the concordance probability as a function of covariates. Under the Clayton copula, the conditional concordance probability has a simple one-to-one correspondence with the copula parameter for different data structures including those subject to independent or dependent censoring and dependent truncation. The proposed method can be used to study how covariates affect the Clayton association parameter without specifying marginal regression models. Asymptotic properties of the proposed estimators are derived and their finite-sample performances are examined via simulations. Finally, for illustration, we apply the proposed method to analyze a bone marrow transplant data set.

Original languageEnglish
Pages (from-to)42-74
Number of pages33
JournalLifetime Data Analysis
Volume21
Issue number1
DOIs
StatePublished - 1 Jan 2013

Keywords

  • Clayton copula
  • Dependent censoring
  • Dependent truncation
  • Multivariate local polynomial regression
  • Non-informative missing data

Fingerprint Dive into the research topics of 'Local linear estimation of concordance probability with application to covariate effects models on association for bivariate failure-time data'. Together they form a unique fingerprint.

Cite this