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

The functional generalized linear model unifies various statistical models, including functional linear regression, functional logistic regression for binary response, and functional Poisson regression for count data under one framework. We will discuss both prediction and slope estimation by functional principal components analysis. Optimal convergence rates are obtained via an asymptotic equivalence result. An application of functional logistic regression to predict US recessions will be discussed, which seems to outperform some classical predictors in economics.

Harrison Zhou received his Ph.D. in 2004 from Cornell University. He is now an associate professor at Yale University. He is the winner of the 2009 Noether Young Scholar Award and a 2007 NSF CAREER Award.