Cluster Analysis of Elemental Composition of Personal, Indoor, and Outdoor PM2.5 Mass Concentrations

• Activity Number: 254
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Jenny Chen*+ and Marepalli Rao and Patrick H. Ryan
• Companies: University of Cincinnati and University of Cincinnati and Cincinnati Children's Hospital Medical Center
• Keywords: k-means clustering ; mclust analysis ; correlation analysis ; heat map

Abstract

The Relationships of Indoor, Outdoor, and Personal Air (RIOPA) study has collected measurements of the elemental composition of PM2.5 in indoor, outdoor, and personal air samples. The samples are obtained from Los Angeles (CA), Houston (TX), and Elizabeth (NJ) on the mass concentrations of 36 elements ranging from Ag (silver) to Zr (zirconium) from about 70 subjects in each city. Each subject in each city provides a data vector of 108 measurements. The goal was to relate personal air samples of the elements to indoor and outdoor air samples and other covariates. Before proceeding with this objective, we embarked on dimensionality reduction, which is the main objective of this presentation. We perform a number of cluster analysis methods on each of the sample types (indoor, outdoor, and personal) data. It emerges that elements mainly of crustal origin and elements of anthropogenic sources band together separately in the clusters of outdoor samples. In addition, elements associated with vehicular emissions cluster together.