Re: type 3 patch

[John Darrington]

On Wed, Jun 22, 2011 at 02:42:51PM -0400, Jason Stover wrote:
     The patch adding type 3 sums of squares to GLM is attached.  Notice it
     includes John's patch to categoricals.c, which he mentioned will
     introduce a bug in ONEWAY.
     
     Comments welcome.
     

I'm following the latin-squares example at 
http://ssnds.uwo.ca/statsexamples/spssanova/latinsquareresults.html

When I run that dataset, the ssq vector returned from get_ssq is:

66.5633
78.8692
45.3287
29.0483
.
.
.

The Type III sums of squares given in the example are:


Source                  Type III SSQ
Corrected Model         263.064
Intercept               815.102
A                       78.869 
B                       28.599 
C                       155.596
Error                   66.563 
Total                   1144.730 
Corrected Total         329.628


So we can see that the first two items in the vector correspond to the SSQ for 
"Error" and "A" respectively.  The other items don't seem to match anything.  
How do we get the sums of squares for "B" and "C" ?

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