You are here

A regression model for longitudinal change in the presence of measurement error.

TitleA regression model for longitudinal change in the presence of measurement error.
Publication TypeJournal Article
Year of Publication2002
AuthorsN Yanez, D, Kronmal, RA, Shemanski, LR, Psaty, BM
Corporate/Institutional AuthorsCardiovascular Health Study,
JournalAnn Epidemiol
Volume12
Issue1
Pagination34-8
Date Published2002 Jan
ISSN1047-2797
KeywordsAged, Bias, Coronary Disease, Humans, Lipoproteins, Models, Statistical, Regression Analysis, Risk Factors
Abstract<p><b>PURPOSE: </b>The analysis of change in measured variables has become quite popular in studies where data are collected repeatedly over time. The authors describe some of the potential pitfalls in the analysis of change when the variable for change is measured with error. They show that regression analysis is often biased, possibly leading to erroneous results.</p><p><b>METHODS: </b>A simple method to correct for measurement error bias in regression models that model change is presented.</p><p><b>RESULTS AND CONCLUSIONS: </b>The two examples illustrate how measurement error can adversely affect an analysis. The bias-corrected approach yields valid results.</p>
DOI10.1016/s1047-2797(01)00280-0
Alternate JournalAnn Epidemiol
PubMed ID11750238
Grant ListN01 85079 / / PHS HHS / United States
N01 85086 / / PHS HHS / United States