Dear Jen,

Howdy   

Although it's a fairly straight-forward matter to estimate different sets of nucleotide rates for different partitions of a data set, the rates are confounded with time.    Branch lengths are therefore expressed in the expected number of A<->G substitutions (hence the 1).  However, to get a useful comparison of AG rates, you could compare the total length of trees estimated for each partition as a proximate measure of variation among partitions in substitution rates, which assumes that those partitions (residing on a common sequence) are beholden to the same evolutionary history.

You can obtain the inferred total counts of transitions and transverions by extracting the reconstructed ancestral sequences.  I've done this for codon substitutions, but I don't have a batch file that handles this on a nucleotide level.  I could probably throw something together, though.  Lemme have a look.

Hasta,
- Art.