Class 18: Animated Sensitivity of Robust Estimators

Review from previous class

Unix Tools, Again

If you're having troubles with your windows (like no scroll bars, ugly colors, etc), take a look at the .Xresources file in your home directory. Here is how mine starts: 
	xterm*background:	white	
	xterm*foreground:	black
	xterm*scrollBar:	true
	xterm*saveLines:	200
	xterm*scrollLines:	5
	xterm*geometry:		80x20
The settings shown here make the xterm window use black text on a white background with a scroll bar. Notice that if you make changes to your own .Xresourses file, the changes won't have any effect until the next time you log in (or take a look at the man page for the xrdb command).

Animating a sensitivity analysis of a robust estimator

Lisp script for today's class
Class18.lsp

Robust estimation
Robust statistics offer a compromise of the efficiency of the mean for Gaussian data with the resistance to outliers of a median. The trick is to find the best of both worlds.

We will also write a small program to estimate a robust statistic using the method of iteratively reweighted least squares.

Animations
The LispStat support for graphics makes it a relatively simple task to build a window that demonstrates the sensitivity of a location estimator to adding an observation.

Next time

Some smoothers like loess make use of robust methods. We'll start looking these more carefully after we continue with our treatment of interactive computing and animations.