Typifying developmental trajectories - a decision making perspective
Alexander von Eye, Eun‐Young Mun, Alka Indurkhya · 2004
Developmental trajectories are defined as curves of repeated observations. Individuals may differ in the starting point, the degree of acceleration or deceleration, the timing of acceleration or deceleration, overall shape, elevation, and scatter of curves. This article dis-cusses methods for typifying developmental trajectories. Two groups of methods are consid-ered. The first group involves assigning individuals to a priori existing trajectories and count-ing the number of individuals that reflect natural groupings of trajectories based on categori-cal classifications, using Configural Frequency Analysis (CFA). The second method involves employing methods of cluster analysis. When selecting a method of cluster analysis, the following ten cluster characteristics need to be considered: (1) disjoint vs. overlapping clus-ters; (2) hierarchical vs. non-hierarchical clustering; (3) agglomerative vs. divisive cluster-ing; (4) exhaustive vs. selective classification; (5) stochastic vs. deterministic clustering; (6) clustering based on correlation vs. distance measure; (7) convex vs. non convex clusters; (8) clustering based on symmetric vs. asymmetric measure; (9) monothetic vs. polythetic classi-fication, and (10) manifest versus latent variable clustering. A review of clustering methods