Nate Silver has an interesting article
which you may have missed from 2010, where he explains per his regression analysis how converts polls of all adults, registered voters and likely voters into an apple to apples comparison (Dems do 7 points better with all adult polls, and 4 points better with registered voters). The Rasmussen "house effect" is 2 to 3 points. He also says PVI isn't as good as party identification. I just thought that you poll junkies might be interested. Whether the same adjustment percentages obtain this year, is of course an issue.