(For exampe, ?expected u? for AL dicodons is 0.833... because there are 24 AL dicodons in all, and 2/24 = 1/12 = 0.8333...)
To my way of thinking, this is another improvement in the new design I proposed in my post of 10/8 at 9:40: since ?expected u? per dipeptide is now the same for the S and C subsets, we have now eliminated one difference between S and C that might have introduced an interfering variable into the analysis.
(In this regard please note that S and C subsets now have the same cardinalities and ?expected u?s? for dicodons set R1 as well as dicodon set 1, and also for dicodon sets 2, 3, R1, R2, and R3.)