Fall 2012: syllabus

Homework sets: homework 1, homework 2, homework 3, homework 4

Final examination:

Notes:

1. 1.Density and distribution estimation
   

2. 2.Regression (as of 10/14)
   

3. 3.Additive regression, Additive regression 2, Additive regression 3
   

4. 4.Notes on function spaces: set 1, set 2, set 3
   

5. 5.Notes on Splines: set1, set2
   

6. 6.Paper on nonparametric regression estimation with parametric covariance (Martins-Filho and Yao (2009)). Slides are here.
   

IMPORTANT ANNOUNCEMENTS:

1. 1.Homework 4 and the research project are due December 17 by 6:00 PM.
   

Data sets: wheatcbot09042012.txt

Codes:

1. 1.kernels.m is a small code with kernels and plots we saw in class.
   

2. 2.rosenkernel.m is a code that implements the Rosenblatt estimator. The code generates data from a gaussian density that you control the parameters. The bandwidth is a given function of n. A plot of the true and estimated densities is produced. You should experiment with different values of c, insert different kernels in the code, generate different data, etc. Play around and see what happens and what you learn experimentally.
   

3. 3. monte_carlo_1.m is a code that implements the small Monte Carlo study requested on question 5 of the first homework set. Use it as an initial template for future simulations.  The code calls some outside functions. They are mixnormbimodal.m and mixnormtrimodal.m.
   

4. 4.nw_example.m is a simple code that implements an univariate Nadaraya-Watson regression estimator and graphs it together with the true regression for two different data generating process. The OLS estimator is also graphed. The code calls the function unw.m
   

5. 5.ugcv.m is the cross validation function for density estimation based on a gaussian kernel. It is called as an input to the function fminsearch in rosenkernel_cv.m.
   

6. 6.mvllin.m performs multivariate local linear estimation. mnw.m performs multivariate Nadaraya-Watson estimation.  Use these modes to obtain jackknifed versions to be used in homework 3.
   

7. 7.spline2.m implements univariate spline estimation based on a polynomial splines of order k and B-splines of order 1. The two functions that are called to construct the bases for the spline space are pspline.m and bspline.m.
   

8. 8.spline1.m implements the spline backfitted kernel estimator of Wang and Yang (2007, Annals of Statistics). A number of functions that perform auxiliary calculations are called.  They include: bspline.m, unw.m, cest.m, dmean.m, nspline