American Statistical Association New York City Metropolitan Area Chapter Mailman School of Public Health Columbia University Department of Biostatistics Colloquium

Applications of cluster analysis have been proliferating, especially in the life sciences. In this talk, I will begin with an overview of clustering methods and then briefly characterize and illustrate current practice. A critical challenge in cluster analysis is to figure out how best to scale or weight the variables involved. One can distinguish between methods that aim to equalize the impact of each variable and those that attempt to highlight variables that are most promising for finding clusters in data. In each of these two situations, there are both univariate and multivariate approaches to the problem. I will describe several of them and show results from experiments designed to assess their effectiveness.