Some additional example metrics.  These are from the RSM and pascal
(combined) projects.

This example predicts changes (errors) by using (halstead) volume,
comments, functions (number of functions in file) and returns
(total returns).  There are 490 files (data points).  The data contains
a number of significant outliers.  For every variable, the standard
deviation is greater than the mean.

Note that I included function count rather than mccabe.  This
seemed to be a more significant variable.

The regression equation is:

changes	=   0.00164 volume  -  0.1097comments  -  0.2515 functions  
	  + 0.07432 returns  +  3.18609

R-Squared is .5818

Note that, once again we have a counter-intuitive variable.  This time
functions varies inversely with changes.  (That is, more functions in
a file imply fewer changes.)  Strange; I speculate that since file size
is explained by the volume variable, many functions in a file means small
functions.  Small functions typically contain fewer errors.

The correlation matrix is:

changes		1.0000
volume		0.6630	1.0000
comments	0.2152	0.7068	1.0000
functions	0.2394	0.6311	0.7528	1.0000
returns		0.6879	0.7087	0.1825	0.3079	1.0000


The t test results are:

volume		10.5995
comments	 5.0503
functions	 1.8426
returns		 3.8045
