|
Activity Number:
|
58
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 3, 2008 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #300615 |
|
Title:
|
Modeling Change in Longitudinal Studies: Use Age Only or Initial Age and Time?
|
|
Author(s):
|
Christopher H. Morrell*+ and Larry J. Brant and Luigi Ferrucci
|
|
Companies:
|
Loyola College in Maryland and National Institute on Aging and National Institute on Aging
|
|
Address:
|
Mathematical Sciences Department, Baltimore, MD, 21210,
|
|
Keywords:
|
Mixed-Effects Models ; Multilevel modeling ; Observational study ; Recruitment bias ; Regression
|
|
Abstract:
|
Since many epidemiological studies involve the study of individuals of different ages over time, it often becomes necessary to distinguish between and estimate both longitudinal and cross-sectional differences. This paper examines how the choice of age and time in modeling longitudinal data can affect the results. In particular, age can be decomposed into two components: age at entry into the study (first age) and follow-up time. The implication of using age or first age and time is described for a number of possible linear mixed-effects models that may be used to describe the longitudinal data. The two approaches are illustrated using a number of different examples of data taken from the Baltimore Longitudinal Study of Aging (BLSA). The examples illustrate that the added flexibility provided by the first age and time approach is usually necessary to adequately describe the data.
|
