2009-07-25 07:10:08 UTC
Hi Jay,
I think you are right because we are looking for the largest of tr1, tr2 or tr3. The first test finds the case where tr1 is the largest so in the second test we only need to test if tr2 or tr3 is the largest.
In the end I didn't put this version on the web page for the reasons that Ethan explained, however this method may show more explicitly what is happening? I would like to find out more about the significance of the trace of a matrix in this context (does a higher trace imply greater accuracy?).
Martin