Correlation Matrix Pseudocolor Map Function
What it does
The function corrmap.m displays a pseudocolor
map of the correlation matrix for a input data set. This much
would be easy, but it can also reorder the variables so that they
are grouped by how correlated they are with each other. A modified
k-nearest neighbor algorithm is used to reorder the variables.
An example of its use is shown below. Here the data file plsdata
was loaded into the MATLAB workspace. The plsdata file contains
data from a slurry-fed ceramic melter process for solidifying
the reprocessing wastes from nuclear fuels (yes, I know that this
example is a bit unusual). The variable xblock1 contains 300 observations
of the 20 temperatures measured within the melter. The variable
yblock1 contains the molten glass level at each time. The MATLAB
command window shows the functions that were called to develop
a correlation map of the 20 temperatures plus the level. (Note:
it is generally much easier to write a small script to make up
labels rather than writing it on the command line.) When the function
is called, it produces the plot shown in the second figure. (Temperatures
are labeled L for left side of the melter, R for right, numbered
from 1 at the bottom to 0 at the top, the level is LV.) Note how
the variables have been arranged so that the highly correlated
ones are near each other in the figure producing a psuedocolor
map where the the high correlations are mainly along the diagonal.
The function can also be used without variable re-ordering, as shown here.
Example of Command Window when using CORRMAP.
Example of Figure Produced by CORRMAP with Variable Reordering.
Requirements for running corrmap
- MATLAB 5.0
- No other toolboxes required
Barry M. Wise
Eigenvector Research, Inc.