Individual aging processes are characterized by complex patterns of interrelated factors including behavioral and psychological factors. In addition, the aging population is possibly heterogeneous. Currently used methods either address complex relations between multiple factors and ignore the possibility of population heterogeneity, or are suitable to assess differential development for multiple types of individuals only with respect to a single factor. Ignoring population heterogeneity when applying regression techniques or exploratory factor analysis may lead to biased results. Although heterogeneity may occur with respect to single factors, more interesting cases of heterogeneity concern the interrelations of multiple factors, which may have a differential pattern of effects on health for different types of individuals. Novel methods for heterogeneous populations have been proposed, and extensions of these models have the potential to assess complex interrelations between multiple factors at a single time point, or in a longitudinal context. It is currently unknown under which conditions (i.e., sample size, data quality, model complexity) the models produce reliable results, or whether the correct model would be selected in a comparison of alternative models that have conceptually different interpretations.
The aim of this project is to assess the potential and illustrate the application of these novel methods. Using extensive simulation studies, the performance of these models will be investigated both in a cross sectional and a longitudinal context. Special focus of the simulations will be on sample size requirements, different item response formats, and different model complexity. The simulations are expected to provide guidelines regarding the feasibility of applying these methods to empirical data. The methods will be applied to existing data of the Notre Dame Longitudinal Study of Aging. The data includes questionnaires of positive and negative affect, life satisfaction, life events, stress, and health. The goal of the analysis is to assess interrelations of these factors both cross-sectionally and longitudinally while simultaneously addressing the possibility that the population is heterogeneous. Relevance: Identifying different types of individuals is a preliminary to develop type-specific interventions to improve health in the elderly. New statistical methods to identify complex typologies have been proposed but require thorough testing prior to applications to assess differential effects of psychological factors on health. ? ?
| Tueller, Stephen; Lubke, Gitta (2010) Evaluation of structural equation mixture models Parameter estimates and correct class assignment. Struct Equ Modeling 17:165-192 |
| Lubke, Gitta H; Hudziak, James J; Derks, Eske M et al. (2009) Maternal ratings of attention problems in ADHD: evidence for the existence of a continuum. J Am Acad Child Adolesc Psychiatry 48:1085-93 |
| Lubke, Gitta; Neale, Michael (2008) Distinguishing between latent classes and continuous factors with categorical outcomes: Class invariance of parameters of factor mixture models. Multivariate Behav Res 43:592-620 |
