George: (this is not a slime throwing post btw...GWB might win.  I have no illusions here)

I'm just lookin' at the Rassmussen Report daily polls and GWB has been leading in that poll since early Feb.  where other polls have shown a lot of back and fourth and even some other polls have shown Kerry in front.  So, this lack of "inter-rater reliability" shows that the polls, while a heckuvalotta fun are really no way to predict.  (inter rater reliability=that seperate sources studying the same thing find get the same results.)

Things one must consider is Population (N).  The larger the N, the more valid the statistics (generally).  This is a daily poll of 1500 people.  For a daily poll, that's really not bad as compared to around 500 that some others are doing.  The webpage for this one doesn't go into how data is collected but given over 100 years experience in consumer index polling, political polling, etc. I sincerely doubt they have too much of an agenda regarding politics...they're just doin' their job.  But what I want to know, is it a longitudinal study or cross section study. 

A cross section takes data about something at one point at time.  Day by day, they check different people and that's how they come up with their numbers.

A longitudinal study follows individuals over time.  I think this would be the best way to see if people's minds are changing.  That way, we could see if ad's are working by swinging undecideds or actually converting people.

A cross sectional study isn't quite so good because you only get opinion from that moment in time from THOSE people.  So, if you're consistently randomly selecting subjects, it doesn't take into account those just randomly saying the first name that comes to mind, it doesn't account for selecting from populations that lean one way or another and it doesn't track change. (go to the rasmussen website and look at the difference between northeast states and say...KANSAS which is notoriously right leaning) So, at the moment that data was taken, the results are probably close to the "truth." But the more time that passes, that data gets stale.  But still, I think the Rasmussen folks "have it goin' on" in that they refresh their data, they go get it rather than soliciting responses (which ALWAYS skews political polling) so those numbers feel pretty good for now.

Margin of error must be considered.  IF the standard deviation of the research tool is greater than the difference between the two sets of data, it could mean that the data is skewed and could be completely wrong.  With them being so close, even a deviation of 10 percent could throw the entire thing off.

Statistics never lie, but liars use statistics:  What we know is that we don't know what the future holds.  I think if you look at polls in October, you'll have a better chance of depicting the future.  Right now, the margin is too close, there's still over 7 months until the election and a lot will happen. 

So, happy polling.  And