507 – Innovative Solutions to Global Problems: Applications from Statistical Practices
Methods for Classifying Changes in Bacterial Prevalence Over Time
Raymond G. Hoffmann
Medical College of Wisconsin
Thomas J. Hoffmann
University of California at San Francisco
Sandra McLellan
University of Wisconsin-Milwaukee
Jessica VandeWalle
University of Wisconsin-Milwaukee
Information about the prevalence of bacterial taxa in water samples can be determined via next generation sequencing. Samples were taken from two different, but related, sites in Lake Michigan over a three year period. After aggregation of similar taxa there were 22 time series of bacterial prevalence x two sites. The goal was to identify the taxa that had similar temporal patterns. Wavelet analysis after filtering to reduce sampling noise was used to determine the temporal characteristics of each of the time series because of the irregular time measurements and the jagged shape of the prevalence curve. Comparisons of clustering of the wavelet coefficients, the coefficients of a harmonic regression model and classification and regression tree (CART) analysis with Random Forests are made to identify the stability and utility of the classifications across different cluster thresholds.
