"Oh Ken, you can use facts to prove anything.  33% of all people know that." - Homer Simpson

In my college statistics class, I remember there are two cardinal rules of statistics: (i) you should do all you can to reduce bias and (ii) guard yourself against your own bias.  Basically, what my dear professor was getting at was to avoid extrapolating or editorializing on data.  Chances were that your collection method was already flawed beyond the point of hope, but you might still pull some gem out of a data set by not making things worse.

Lately, I've seen no end to the surplus of both sides of this primary breaking rule two.  Compounded with the sampling error that pollsters inherently have, this is a dangerous combination.  Let's look at some recent polls:

SurveyUSA Poll:

Obama +10 in NC
Obama 50%, Clinton 40%
Sample: 580 LV
MoE: +/- 4.2%

Poll predicts relatively good chance of an Obama victory.  For our purposes, it's safe to say that from this data set.  But wait!  That's just this data set and - while it is following conventional sampling rules - it's not necessarily true; New Hampshire demonstrated that fairly clearly that otherwise good polls can miserably miss the actual result.  Our good friend sampling error has almost everything to do with that.  Is this poll most likely correct?  Yes but not necessarily.  We can doubt what the definition of a "likely voter" is and that participants were truly chosen randomly, but this approaches absurdity real quick.

With this in mind, let us take a look at preliminary GE polls - Obama vs. McCain, Clinton vs. McCain.  Already, we have the traditional doubts.  But furthermore, we have a brand new problem - can polls taken in February be used to predict an outcome in November?  I think the answer to everyone is obviously no.  The intricacies of this election far exceed what a simple "If the general election were held today..." question can encapsulate.  The steps required to go from point A were a candidate is ahead in general election polls in a primary to point B were that candidate is necessarily more electable than a rival require much more information than what one poll can present and is simple extrapolation.

Now, as a final point to this diary, keep in mind that pollsters are in the business of extrapolation and, as our good friend Mr. Zogby demonstrated, any one of them can be biased.  So, in short, I would advise both sides of the primary to state your opinions but don't pretend to have hard proof - it's like the front pagers remind us, "only actual numbers matter."