GLM fixed ! (nearly)

[John Darrington]

The problem turned out to be in the categorical encodings (again).
I've pushed several commits which now means we have a full factorial
anova GLM with Types I, II and III sums of squares.

I've done some reasonably comprehensive tests, and can only see 
one problem:  When presented with an unbalanced design and Type III
is requested the Intercept value doesn't agree with SPSS.

Now we're calculating the Intercept sums of squares simply as 
(mean^2 * N). SPSS is clearly doing something different.  Quite what,
I don't know.

If someone can tell me how to get this final bit of the puzzle solved,
I think we'll have made a big step forward.

J'

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

signature.asc
Description: Digital signature