Polling organization, dates, sample size and LV=likely voters, Clinton%, Obama% and the undecided%.
|PPP (D)||5/3-4/08||851 LV||51||46||3|
The average of these gives a 5.5% difference, with 4.7% undecided. Allocating those as if they will split like everyone else gives 5.8%. If they all go for Clinton, then we get to a 10% difference.
One of several interesting things about this is that while the standard deviation of the Obama percentages is 1.4, the standard deviation of the Clinton percentages is 4.3. The standard deviation of the undecided percentages is 3.1. Basically, all those undecideds go over to Clinton or not, depending on specific details of the question phrasing, and how the organization balances their sample to the statewide demographics.
I'll drop my prediction from Clinton by 14% in the previous post to Clinton by 10%. Same standard deviation as before, 4%.