Mitt will win quite easily in Arizona (no need to go into percentages b/c it's winner-take-all).
As for Michigan, it really is way too close to call. In theory, Santorum should have the edge b/c of the makeup of the state, and Democrats playing games (which was a big problem for Bush in 2000), but home-state bumps are tough to really quantify, even in polling. I have to think that if he gets more than 39% (Mitt's total in 2008), he's probably gold, but who knows.
This guesstimate (i.e. Al-style prediction) appears to be about right.