Dr. Kesheng Wang, associate professor in the department of biostatistics and epidemiology in the ETSU College of Public Health, authored an opinion article in the Journal of Biometrics & Biostatistics. The article, “Linear and Non-Linear Mixed Models in Longitudinal Studies and Complex Survey Data” examines the development and application of mixed models in analysis of correlated data.
[Photo: Dr. Kesheng Wang]
Correlated data are fairly common in health and social sciences. For example, clustered data arise when subjects are nested in clusters such as classrooms, hospitals, and neighborhoods; while longitudinal data result from multiple measures for the same subject over long period of time. Observations for the same cluster/subject are likely to be correlated (non-independent).
Mixed models are commonly used to deal with correlated data or hierarchical data in health and social sciences. These models, also known as multilevel or hierarchical models, include both fixed and random effects. Linear mixed models (LMMs) are extensions of linear regression models, which describe the relationship between a continuous response variable and independent variables, with fixed effects and random effects.
Mixed models have now been extended to analyze multiple outcomes in longitudinal study. The models are useful because they can incorporate random effects for single outcome and also for shared or separate random effects for multiple outcomes. Acknowledging random effects can increase the power in estimation of the fixed effects, avoid the serious inflation of the type I error rates, and prevent false positive results.
The Journal of Biometrics & Biostatistics is a broad-based journal and publishes current research on subjects including biometrics, hypothesis testing, statistical methods, and clinical trials.