I’ve been playing with my calculator, and my best guesstimate is that Ras is currently understating McCain’s support by somewhere between .85 of a point and one full point due to the rapidly changing partisan identification numbers (i.e., not taking into account sampling error or any other way that error can be introduced.)
This is not to imply in any way that Ras is TRYING to understate McCain’s numbers. To the contrary, Ras bends over backwards to get it right, but knowing when one is right is impossible in polling, at least until November 4th.
Rolling averages are a time-honored method of reducing statistical noise. That is why he is using them.
To Rasmussen’s credit, he goes out of his way to explain how he does what he does and why he does it.
Lots of other folks don’t share how they concoct their brew.
FWIW, RCP’s average of polls is just about where Rasmussen says things are.
You’re thinking that McCain is actually up by about 2-3 in Rasmussen’s numbers. We’ll not see the full effect of the last day of the Pub convention until tomorrow, iirc.