Clustering on the basis of longitudinal data

Area of Science:

• Biometrics
• Longitudinal Data Analysis
• Statistical Computing

Background:

• Analyzing longitudinal data requires methods to quantify differences in individual growth trajectories.
• Estimating polynomial growth curves is a common approach in developmental and health sciences.
• Existing methods may lack user-friendly tools for direct comparison of growth curves.

Purpose of the Study:

• To introduce ZDIST, a menu-driven PC program for calculating distances between polynomial growth curves.
• To provide a computational tool for assessing similarities and differences in longitudinal subject growth.
• To facilitate the application of clustering techniques to growth curve data.

Main Methods:

• Development of a menu-driven PC program named ZDIST.
• Computation of distances based on estimated polynomial growth curves.
• Inclusion of options to use growth velocity and acceleration curves for distance estimation.
• Saving distance matrices in ASCII format for compatibility with other software.

Main Results:

• ZDIST provides a straightforward method for computing inter-individual growth curve distances.
• The program allows flexibility by incorporating raw curve data, velocity, and acceleration estimates.
• Generated distance matrices are compatible with standard clustering algorithms like SYSTAT.

Conclusions:

• ZDIST offers a valuable and accessible tool for researchers analyzing longitudinal growth data.
• The program enhances the ability to perform quantitative comparisons and cluster subjects based on growth patterns.
• Availability of ZDIST supports advancements in the statistical analysis of developmental trajectories.