Salon E
Some Practical Applications of Estimation and Inference with Shape Restrictions (303333)
Xiyue Liao, Colorado State University*Mary C Meyer, Colorado State University
Keywords: monotone, convex, cone projection
Many statistical methods involve the estimation of a function; for example: regression function, dose-response curve, probability function, hazard function, growth curve, variance function. Parametric methods require the user to specify the function up to a few parameters, but often the valid assumptions do not warrant this specification. In this talk we consider the generalized additive model with shape and smoothness assumptions for the various components. A model selection criterion can choose both the variables and the shapes.