About Those Polls ...
Statisticians know that the single most important factor in determining the credibility of a survey is the randomness or lack thereof in the selection of those interviewed.
To be representative, a survey must give every potential interviewee an equal chance of being interviewed. Perfect randomness is impossible to achieve, but it is the goal that must be striven for in order to do credible survey work.
That is why it is critical in evaluating the credibility of anybody's survey to take a look at who was interviewed and their characteristics, or weighting, to determine if the survey results are representative.
Do that with the latest trio of surveys showing President Bush at an all-time low in popularity and you get an idea of how critical representativeness can be. Fortunatley, Ankle Biting Pundits has already done it for you. Go here.