Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
I like how Gallup has corrected to +1 Romney from +5 to +7 Romney right before Sandy. Is it the Sandy effect? Or did Gallup tweak their LV model a bit?
Same thing with Pew, Obama got a bump but its confined to the Northeast. Which is inconsequential in the electoral sense for the most part.
sabato is a legit pollster. im glad he has obama winning.
Sabato is good but other good analysts are saying the opposite which speaks to the close nature of the race. The notion that Obama is a 90% favorite is laughable.
Sabato is good but other good analysts are saying the opposite which speaks to the close nature of the race. The notion that Obama is a 90% favorite is laughable.
What's laughable is your faith in Mitt. You guys will rip Romney to shreds shortly after his concession speech.
Fair enough but most of the underlying polls are flawed, garbage in, garbage out. RCP may have to remove PPP, Marist and even Quinnipiac when the dust settles.
Or they could just include even more polls to get a better snapshot of the electorate. I mean, once your aggregate sample size gets really big (15,000+), it's pretty hard to claim bias. Ohio's a big state, but it's not that big. Polling thousands and thousands of voters in the state should give you a good idea of where the electorate stands. Do you want an aggegate sample size of 25,000? 50,000? 150,000?
Green Onions is correct. Error is not uniformly distributed. Error has a t-distribution which looks a lot like a normal distribution but is thicker at the tails. The error with the sample sizes under consideration it is, for practical purposes though, a normal distribution - the bell curve which many of us are familiar with. This means that the chance of the real outcome is highest at the published number and declines significantly at the margins of the margin of error. In simple terms, it would be a rare event if the actual population value was the sample value plus or minus the margin of error. Values closer to the sample value are more likely and the published value of the sample is the most likely. But, of course, there is always a small, but finite probability that you are correct and that Romney will perform at the margins or even better. Don't count on it, though. I don't have access to Silvers model, but his 14% looks a lot like what that probability would be.
And Gallup (one of the posters to which certain desperate folk cling, like a drowning man clings to a piece of debris in the water) has President Obama at 52% approval rating.
Or they could just include even more polls to get a better snapshot of the electorate. I mean, once your aggregate sample size gets really big (15,000+), it's pretty hard to claim bias. Ohio's a big state, but it's not that big. Polling thousands and thousands of voters in the state should give you a good idea of where the electorate stands. Do you want an aggegate sample size of 25,000? 50,000? 150,000?
So you agree with Gallup and Ras take on Party ID split?
Green Onions is correct. Error is not uniformly distributed. Error has a t-distribution which looks a lot like a normal distribution but is thicker at the tails. The error with the sample sizes under consideration it is, for practical purposes though, a normal distribution - the bell curve which many of us are familiar with. This means that the chance of the real outcome is highest at the published number and declines significantly at the margins of the margin of error. In simple terms, it would be a rare event if the actual population value was the sample value plus or minus the margin of error. Values closer to the sample value are more likely and the published value of the sample is the most likely. But, of course, there is always a small, but finite probability that you are correct and that Romney will perform at the margins or even better. Don't count on it, though.
I'm not disputing the number in the top line is more likely to happen than anything else under the bell (assuming the sample is good which I'd argue they are not) just that it is not statistically significant. A one point lead which translates to about 10 people in a N of 1000 is not much to say one is ahead. It is statistically insignificant.
Also with a sample of 1000 the distribution would be normal not T which is for small samples.
So you agree with Gallup and Ras take on Party ID split?
No (at least not with Ras...not so sure what you mean by Gallup). If I randomly call 1,000 people, 500 identify as Dems, 480 identify as Repubs, and 20 identify as Independents, I can't keep calling around until the number of Dems and Repubs evens out. That's like trying to find out whether people prefer Butter Pecan or Strawberry ice cream and then not accepting the answer people give me ("They just have to prefer butter pecan!")
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.
Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
