Description

Non-parametric regression techniques and concepts such as splines, kernels, regularization, and cross-validation are important for the development and understanding of modern machine/statistical learning tachniques and also extremely useful and flexible tools for data analysis. Students in this course will learn how to use non-parametric techniques to analyze data and how the methods work. The course will focus on practical methods for data analysis, not just theory.

Tentative topics to be covered include:

• Estimating the CDF and statistical functionals.
• Bootstrap and Jacknife
• Kernel smoothing
• Density estimation
• Penalized regression; splines
• Semiparametric and wavelet regression

Prerequisites

Undergraduate course in regression and Stat-6510.

Text

• Semiparametric regression by Ruppert, Wand, and Carroll