1.  Movies that teach about local polynomial smoothing
by J. S. Marron, D. Ruppert, E. K. Smith and G. Conley.
 

Complete discussion of lessons shown by the movies available here can be found in the paper "Motion Picture Analysis of Smoothing", by Marron, Ruppert, Smith and Conley, North Carolina Institute of Statistics, Mimeo Series #2367 ( PDF version (286KB)  /  Postscript Version (1.49 MB) ).

A listing of such movies viewable from here:

Group 1:    Comparison of Bandwidths, Blip Target.
           Movie 1a, Local Linear
           Movie 1b, Local Cubic
           Movie 1c, Local Quintic

Group 2:   Comparison of Bandwidths, Piecewise Target
           Movie 2a, Local Linear
           Movie 2b, Local Cubic
           Movie 2c, Local Quintic

Group 3:    Comparison of Polynomial Degree, 3 Period Sine Target, Low Degree (0,1,2)
           Movie 3a, Small Bandwidth
           Movie 3b, Medium Bandwidth
           Movie 3c, Large Bandwidth

Group 4:    Comparison of Polynomial Degree, 3 Period Sine Target, High Degree (2,3,4)
           Movie 4a, Small Bandwidth
           Movie 4b, Medium Bandwidth
           Movie 4c, Large Bandwidth

Group 5:    Comparison of Kernels( Uniform, Triweight, Gaussian), Blip Target.
           Movie 5a, Ordinary Scaling
           Movie 5b, Canonical Scaling

Group 6:    Comparison of Sample Sizes and Sigmas, Blip Target.
           Movie 6a, Different n
           Movie 6b, Different Sigma
           Movie 6c, Signal to Noise Ratio Fixed

Group 7:    Comparison of X Designs, 2 Period Sine Target.
           Movie 7a, Random Design, degree 2
           Movie 7b, Random Design, degree 3
           Movie 7c, Fixed vs. Random Designs

Group 8:    Comparison of Scedasticity, 3 Period Sine Target.
           Movie 8a, Increasing, Constant and Decreasing
 

These were generated in Matlab, using software that is web available, at:  http://www.unc.edu/depts/statistics/postscript/papers/marron/Movies_Software/.  This software allows easy menu driven creation of the movies posted here.  These Matlab versions looks a little better because MPEG is a bit lossy. There is also menu driven control of some parameters, such as number of frames.  If you are comfortable programming matlab, this software can be adapted to create your own movies, which investigate/illustrate your favorite aspect of local polynomial smoothing.

Back to Movies Table of Contents

Back to Marron's Home Page