Statistics 526 Term 2 1998-99
Smoothing Methods in Regression

JANUARY 5
Check of class's regression background
Quick summary of least squares and logistic regression
Handout on birthweight data
JANUARY 12
Overview of major smoothing methods
Handout on smoothing methods
plus ....
JANUARY 19
Nadaraya-Watson estimate:
definition, asymptotic bias and variance
Lab: playing with the Splus's ksmooth, to understand bias
JANUARY 26
More Nadaraya-Watson estimation: asymptotic bias, variance
Kernels of different orders
Estimating derivatives
Handout: my Splus function myksmooth,
using kernel of order 4
Lab: homework discussion
FEBRUARY 2
Choosing the smoothing parameter: by eye, effective number of parameters,
plug in (first and second generation), cross-validation
Lab: estimating derivatives, growth curve data
Handouts:
Number of parameters,
Plug in choice of h,
Derivatives of growth data
Write-up
(not distributed in class)
Choosing Smoothing Parameter
FEBRUARY 9
Choosing the smoothing parameter: cross-validation
Discussion of beluga whale data
FEBRUARY 23
Simulation results: comparing bandwidth choice - but dated February 9.
handout,
longer version
Local linear methods:
short handout, which was distributed,
and long write-up, which wasn't distributed.
MARCH 2
Loess/locpoly
Multiple Regression Kernel Methods
Generalized Additive Models
Generalized Additive Models: Splus
MARCH 9
Hatmatrix for 2-D loess, bivariate gam (not distributed)
Generalized Additive Models continued
Parametric Model Fitting tex write-up (not distributed)
Model Fitting in Gam
MARCH 16
General Linear Models, Iteratively Reweighted Least Squares
Discussion of data sets: please find a data set and use
gam to decide on an appropriate model.
MARCH 23
More General Linear Models and General Additive Models,
with information in the
write-up, and Splus examples.
MARCH 30
Least squares regression and
Bspline regression.
Smoothing Splines and Penalized Likelihood.
Smoothing spline
examples in Splus.
Wan-lin's presentation on bootstrapping
APRIL 6
More smoothing splines
Peter's neat basis functions, with standard errors for estimates
now available!
WED APRIL 21 1-4 pm
Presentations by Huiying, Isabella and Sijin
|