Describing the natural heterongenity of aging using multilevel regression models

Abstract

Aging has been defined as the process of change that occurs in the individual during the course of time following the early stages of growth and development. While this process occurs in everyone, it varies from person to person. Longitudinal studies have emerged as the only method to study individual change directly and to identify factors associated with that change. Multilevel or mixed-effects regression models have proven to be a useful tool for describing the natural heterogeneity that occurs in studies of aging. These models, along with recent developments in estimation procedures and numerical techniques, have made it possible to estimate in a unified analysis the average rates of change for the study population, as well as individual deviations from these average rates. One type of multilevel models, mixed-effects models, assumes that the correlation among longitudinal measurements for an individual is due to some latent characteristics that give the individual an initial level or rate of change which is higher or lower than average. This paper discusses the application of mixed-effects models using random effects for the estimation of individual differences to aspects of human aging which have been observed over the first 35 years of the ongoing Baltimore Longitudinal Study of Aging.