Abstract:
|
A level set of a density is a set where the density equals a given constant. It has many applications including clustering, classification, anomaly detection and topological data analysis. Level sets can be estimated nonparametrically using a plugin method. In this talk we present confidence regions for level sets based on asymptotic distribution and bootstrap methods. We then make comparisons with existing methods using simulations. This is joint work with Wolfgang Polonik, University of California, Davis.
|