Principle Component Analysis as a Statistical Tool To Investigate the Role of Specific Habitat Variables in Lyme Disease Ecology

• Activity Number: 377
• Type: Contributed
• Date/Time: Wednesday, August 9, 2006 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Haiyan Chen*+
• Companies: Dana-Farber Cancer Institute
• Keywords: principal component analysis ; b. burgdorferi prevalence ; small mammal ; habitat variable

Abstract

To investigate the impact of habitat variables on B. burgdorferi prevalence within reservoir mammals, forest and mammal surveys were conducted at 12 sites in New York between 1998 and 2001. Spearman rank correlations between each of 100 habitat variables collected and average species-adjusted B. burgdorferi infection rates (SAR) of the mammals in a site were calculated, 13 of which had p-value < 0.10 and were selected for a principal component analysis (PCA). Shrub, fern, tree, and water are deterministic factors regarding habitat characteristics in a given area. Backward multiple regression analysis was conducted to further identify the best predictors of the SAR. This study found that PCA is an efficient approach to condense the information contained in a large number of original variables into a smaller set of new composite dimensions with a minimum loss of information.