In a standard syllogism, the two premises of the argument are independent of each other, so all extrinsic warrant being equal, the argument will prove the conclusion to be probable if and only if...
P(Premise 1) * P(Premise 2) > 50%
If the probabilities of each premise happen to be equal, this turns out to be about 71% certainty in each. This calculation obviously turns out much higher for arguments with more than two premises.
But in modus tollendo tollens arguments, "not Q" is far from independent of "P implies Q." Taking a mathematical argument for God's existence (the X would be replaced with some particular degree)...
1) If atheism is true, then the world would not be built on mathematical structures of complexity greater than degree X,
2) The world is built on mathematical structures of complexity greater than degree X,
3) Therefore, atheism is false.
Suppose the opponent agrees with premise 1, with the ~71% certainty required for each premise of the other syllogism. In this argument, in contrast to the other syllogism, premise 2 is not independent of premise 1. If we bring to this argument a 50% prior probability that atheism is true (at least methodologically to keep an open mind), then there seems to be a 50% chance that premise 1 provides a defeater for premise 2, so that the argument given for premise 2 would have to yield much higher certainty than 71%. Am I thinking about this in the right way? Is it impossible (or improper) to evaluate the premises of a modus tollendo tollens argument independently of each other? What are the probability requirements for a good modus tollendo tollens argument?