Graphical model checking with correlated response data

Wei Pan, John E. Connett, Giovanni C. Porzio, Sanford Weisberg

Research output: Contribution to journalArticlepeer-review

3 Scopus citations

Abstract

Correlated response data arise often in biomedical studies. The generalized estimation equation (GEE) approach is widely used in regression analysis for such data. However, there are few methods available to check the adequacy of regression models in GEE. In this paper, a graphical method is proposed based on Cook and Weisberg's marginal model plot. A bootstrap method is applied to obtain the reference band to assess statistical uncertainties in comparing two marginal mean functions. We also propose using the generalized additive model (GAM) in a similar fashion. The proposed two methods are easy to implement by taking advantage of existing smoothing and GAM softwares for independent data. The usefulness of the methodology is demonstrated through application to a correlated binary data set drawn from a clinical trial, the Lung Health Study.

Original languageEnglish (US)
Pages (from-to)2935-2949
Number of pages15
JournalStatistics in Medicine
Volume20
Issue number19
DOIs
StatePublished - Oct 15 2001

Fingerprint Dive into the research topics of 'Graphical model checking with correlated response data'. Together they form a unique fingerprint.

Cite this