Class Notes:  Thursday 11/21/02
 

    -    Check new material on student pages (from Class Home Page)
 
 

    -    Class Lunch:    Meet at McAllister's, 11:00AM, on Tuesday, Nov. 26.
 
 

New Tool:  "Scatterplot Smoothing "

(from Section 3.2 in text)
 

Main Idea:    Highlight Structure in Scatterplots
 
 

Example:  Canadian Earning Power Data

Data stored as .mat file

Load into Matlab using ">> load canearn"

Check read in with ">> whos"     (gives list of variables)
 
 

Make Scatterplot

    -    Hard to see much "structure"
 
 

Make Scatterplot Smooth

    -    Use command ">>  nprSM"

    -    Experiment with vh

    -    Experiment with ndataoverlay
 
 

A look "under the hood":    Local Linear Smoothing

Toy Example:

    -    Main Idea:    "inside window" do a linear fit

    -    Then "slide window along" to trace out curve

    -    Different window widths control "smoothness"

    -    Small window    --->    rough fit

    -    Large window    --->    smooth away important features

    -    Same issues as kernel density estimate

    -    Software gives "usually reasonable" defaults
 
 

Back to Statistics 6D Home Page