next up previous contents
Next: Plotting and charting: Data Up: Plotting and charting Previous: Plotting and charting: 2-D   Contents

Plotting and charting: 2-D statistical plotting

Statistical plots have a lot of variety. Here we illustrate a fairly simple classification problem.

The idea is that we are given a set of training data in 2 classes and with 2 observed (or feature) variables. We assume that the measurements are bivariate Gaussian (normal) distributions in each case, but that the parameters are different in each case. The plot shows simulated data, and 2 of the ellipses indicate the mean, variance and covariance estimated separately for the 2 classes. Based on this a third ellipse has been drawn that passes between the others. A Bayesian interpretation is most straightforward. Using as prior probabilities the relative frequencies of classes in the training set, measurements on this ellipse have equal posterior probability of being in each of the 2 classes. For all other measurements, one class has a higher probability than the other.

ScatClass-small.png
Bivariate 2-class classification. Training data points are shown, along with indicators of their distribution. The ellipse passing between the others divides the space of observation values for which each class has a higher posterior probability than the other. Creation method: Octave, eps output, with title and border added in XFig, ps output.

View figure as small | medium | large | huge image.
Download bzipped-tar file of PDF-format images.


next up previous contents
Next: Plotting and charting: Data Up: Plotting and charting Previous: Plotting and charting: 2-D   Contents
Alex Stark 2003-12-27