BIOSTATISTICS SPRING FORUM 2008

October 15, 2008

by Andrew Lawson (MUSC) and Huafeng Zhou (MERCK)

"EWMA smoothing and Bayesian surveillance of disease maps"

Fast computational speed is important if real-time or near-real-time surveillance of large volumes of spatial data are to processed. With a Bayesian approach it is not clear how to limit the computation when the both the data and parameter size increases with the arrival of new data. In this talk we will outline a novel method for fast updating of spatial information where the temporal effects are not modeled explicitly but are allowed to be smoothed via an exponentially weighted moving average
(EWMA) process. This reduces computation considerably as only a spatial model is fitted at each time point. The talk focuses on derived monitoring measures and their properties (Pr(false alarm), ARL0). Simulation results are discussed and an application to a historical salmonella outbreak in upstate South Carolina is demonstrated.

October 29, 2008

by Jijun Tang
USC, Department of Computer Science and Engineering

"Statistic Methods for Phylogenetic Analysis"

Phylogenetic analysis is to determine the evolutionary relationship among species. To date, the dominant data for phylogenetic reconstruction is molecular data, including DNA or Protein sequence data and genome rearrangement data. In the past several years, statistic methods using Maximum Likelihood and Bayesian approaches have been widely used to infer phylogenies. In this talk, we will review several popular tools and provide some discussions about future researches to improve these methods.