Application of Cluster Analysis to Aerometric Data. Volume I. Part 1: Clustering, Validation, and Classification of Data. Part 2: Investigation and Report of Cluster Analysis

摘要

The calibration and enhancement of Wolfe's NORMIX (normal mixtures) computer program in the National Computing Center of the U.S. Environmental Protection Agency at the Research Triangle Park, NC is documented. The program is available for data clustering, validation, and classification. In Part 1, an historical perspective of the NORMIX algorithm is presented with details of calibration, using the Fisher-Anderson Iris data set. New program options are then introduced with illustrative outputs including clustering features such as variate profiles and bivariate representations. In Part 2, Statistical Analysis Systems (SAS) and Multivariate Iterative K-Means Cluster Analysis (MIKCA) clustering algorithms are applied to Los Angeles Catalyst Study (LACS) data to investigate their performance as a multivariate tool for modeling air pollution data. This volume, Volume I, of a 3-volume series comprises the report for Parts 1 and 2. Parts 3 and 4 are reported in Volumes II and III, respectively.