Generalized semiparametric regression with covariates measured with error

Thomas Kneib, Andreas Brezger, Ciprian M. Crainiceanu

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

We develop generalized semiparametric regression models for exponential family and hazard regression where multiple covariates are measured with error and the functional form of their effects remains unspecified. The main building blocks in our approach are Bayesian penalized splines and Markov chain Monte Carlo simulation techniques. These enable a modular and numerically efficient implementation of Bayesian measurement error correction based on the imputation of true, unobserved covariate values. We investigate the performance of the proposed correction in simulations and an epidemiological study where the duration time to detection of heart failure is related to kidney function and systolic blood pressure.

Original languageEnglish (US)
Title of host publicationStatistical Modelling and Regression Structures
Subtitle of host publicationFestschrift in Honour of Ludwig Fahrmeir
PublisherPhysica-Verlag HD
Pages133-154
Number of pages22
ISBN (Print)9783790824124
DOIs
StatePublished - Dec 1 2010

Keywords

  • MCMC
  • additive hazard regression
  • generalized additive models
  • measurement error correction
  • penalized splines

ASJC Scopus subject areas

  • Mathematics(all)

Fingerprint Dive into the research topics of 'Generalized semiparametric regression with covariates measured with error'. Together they form a unique fingerprint.

Cite this