Clustering Methods for Linear Mixed Models

This dataset will be used for two purposes: 1) to explore novel statistical methods for clustering longitudinal data; 2) to satisfy the requirements for a longitudinal data analysis course in the Colorado School of Public Health, BIOS 6643. Dr. Colborn and Dr. Juarez-Colunga will act as mentors on this project, and the goal is to publish the related work in a statistical methods journal. Both mentors are PCRC members and DISC statisticians.

Traditional methods of multi-level linear mixed model often include a “organization” as highest level such as schools, medical centers, or hospitals. Although some people in these clusters may share similar unmeasured characteristics (e.g. socioeconomic status, distance to care, food insecurity/access), there may be other features that would make them similar to groups other than their organization or institution.

Here, I describe “Clustering LMM” a new method to generate a highest level for a multi-level linear mixed model, by reassigning people into different clusters based on demographic and socioeconomic variables available in the data. Below, I describe the potential benefits of this method.