STAT 546: Nonparametric Models in Statistics

This course is offered once every two years. 

Time: 12:40pm TR at Snedecor 321.

First lecture is on Tuesday, September 1st.

NO LECTURE IN THE FISRT WEEK.

Course Outline:  The course is focused on smoothing techniques for estimating density, regression and other functional curves without a parametric family of models.  Smoothing methods, together with the Bootstrap and perhaps the Empirical Likelihood are the key members of modern nonparametric statistical methods.  Thanks to the availability of modern computational and graphical tools and the increases in the amount of data at our disposal, it has become feasible to move away from the classical parametric models.  When we are using nonparametric methods in exploratory data analysis or model building and inference, we are "letting data to speak for themselves."  In the course, we will discuss how to set up a nonparametric model, how to choose the amount of smoothing, and how to evaluate the resulting fits.

The method we will cover extensively is the Kernel Method.  I have been working on nonparametric curve estimation for many years, and  will share the experience with the students.  The course will consists of projects where students have the opportunity to practice nonparametric estimation techniques based on their own data of interest.  The objective is to enable students to understand how and why these  techniques work and know how to implement them in practice.

Credits: 3  Pre-requisites: Stat 511 and Stat 542, or consent of instructor.

Assessement: 40% from one project  and 20% from Assignments, 40% Final Exam. 

Recommended Text:

Simonoff, J. S. (1996). Smoothing Methods in Statistics. Springer.

Other References

Silverman, B. (1986). Density Estimation for Statistics and Data Analysis. Chapman and Hall.

Hardle, W. (1990). Applied Nonparametric Regression. Cambridge University Press.

Fan, J. and I. Gijbels (1996) Local Polynomial Smoothing. Chapman and Hall.