Posted on 10/11/2004 6:33:29 AM PDT by big time major leaguer
That about covers it
in 2000 his "accuracy" was within 3 points of other polls. There was an article on his site that showed how supposedly his poll was more accurate than others.. but when you examined it, it was within 3 points of all the other polls.
This poll that is being shown right now has an 8 point difference. Someone mentioned being wrong in 2002.. i would like more information on that.
I do find it hard to believe that sKerry is up by 3 in any poll.
Which reputable polls? Rasmussen and ABC news/Washington post?
I also did not recieve any email notification or anything about the poll.
However, there is a ironic point to this newest zogby poll. He finds that bush now leads among young voters. Wouldn't be ironic if all these campus rock band tours actually turn people off to the democrats? Wouldn't it be ironic if the campus crowd now feels deceived by the draft rumors the dems floated and there's a backlash? Wouldn't it be just the sweetest thing in the world if the campus GOTV efforts by the dems turn out more bush votes than kerry?
yes, in 2002 and now he used the 2000 turnout model to weight his results.
In 2000 republican turnout was depressed due to the last minute dui scandal.
In 2002 he was wrong in 29% of the races he called - nearly twice the rate (of 15%) of other polling companies.
So, Zogby underestimated Bush support in 2000 by a full 2%? Enough said, I guess.
-T
When is the next round of the "major" pollsters coming out. Seems like after the first debate, AP, Newsweek, Time, Gallup all rushed out with polls showing Kerry "surging". Saw very few majors over the weekend. Tracking polls, with the exception Zogby look positive.
That stated uncertainty is based upon sampling error. Modelling error is not reflected in the stated uncertainty.
Samples are subject to variations due to randomness, and are therefore mathematically predictable. Multiple polls with only random errors will refine the reults and can be combined to increase accuracy.
Modelling error is due to wrong assumptions, and are just wrong. Doing more polls with modelling errors do not increase accuracy. A well-done poll with a modelling error is measuring something other than what is intended.
Yes, since the election is a state-by-state election, it is better to measure state-by-state results, weighted in accordance with the electoral college, than national numbers which are only generally correlative.
I don't think this is accurate.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.