A Small-Scale Investigation of Statistics for Determining the Number of Clusters in a Data Base.

摘要

In construction of a predictive mathematical model of impact acceleration injury, changes in evoked potential response may serve to provide important information. In attempting to determine whether a number of different types of changes exist, cluster analysis may be used. This report describes a limited Monte Carlo examination of four statistics (used in conjunction with Ward's hierarchical clustering algorithm) for identifying the number of clusters built into a simulated ten-dimensional data space. (Author)