PDA

View Full Version : They need new election tea leaves



SanJoaquinSooner
2/6/2008, 03:03 PM
Apparently, polling is more like reading tea leaves these days.

Final Predictions for California Democratic Primary:


Reuters/CSpan/Zogby 02/03 - 02/04 n=895 49% 36% Obama +13.0

SurveyUSA 02/03 - 02/04 n=872 42% 52% Clinton +10.0


Reuters/CSpan/Zogby predicted a 13% victory by Obama in California.

SurveyUSA, using the same sample size and polling on the very same days nailed it: Clinton 52% to Obama 42% - exactly how it turned out.

Whoever designed the sampling procedures for Zogby should be fired.... a swing that large can't be explained by random error.

sooner_born_1960
2/6/2008, 03:11 PM
Pollsters make million putting their crap out. How dare you criticize the results.

Frozen Sooner
2/6/2008, 03:14 PM
You know, I can't recall the last time I saw a Zogby poll that was correct. The guy called one crazy election correctly like 8 years ago and everyone thinks he's the master of polling.

I think there may be something crazy with his methodology. I'm no statistician, certainly, but when polling data is consistently wrong like that, there's something going on.

Ike
2/6/2008, 03:14 PM
Apparently, polling is more like reading tea leaves these days.

Final Predictions for California Democratic Primary:


Reuters/CSpan/Zogby 02/03 - 02/04 n=895 49% 36% Obama +13.0

SurveyUSA 02/03 - 02/04 n=872 42% 52% Clinton +10.0


Reuters/CSpan/Zogby predicted a 13% victory by Obama in California.

SurveyUSA, using the same sample size and polling on the very same days nailed it: Clinton 52% to Obama 42% - exactly how it turned out.

Whoever designed the sampling procedures for Zogby should be fired.... a swing that large can't be explained by random error.



Ummmm....yes it can.


Question: How many polls have there been in the history of the country? Maybe we can even narrow it down and ask "How many political polls are taken in a year"?

With 1000 respondents the sampling error is ± 3%, right?

That means that 68% of the time, the population will vote within ± 3% of the numbers the polls find.

95% of the time, the population will vote within ±6%

About 1 in a million of those polls will miss by close to 20%.

Think there have been a million polls yet?


Now, if they are consistently off, then yeah, they probably have a systematic flaw in their sampling or their weighting. But with so many polls conducted so often in the 24 hour news cycle world we live in, sampling error can have significant effects.

Frozen Sooner
2/6/2008, 03:17 PM
Ike with the explanation of what sampling error is FTW!

Seriously, though, Zogby is pretty consistently out there.

Ike
2/6/2008, 03:23 PM
Ike with the explanation of what sampling error is FTW!

Seriously, though, Zogby is pretty consistently out there.
Yeah, I wouldn't know. Polls don't mean squat to me, mainly because I really don't care that much.

olevetonahill
2/6/2008, 03:33 PM
Ummmm....yes it can.


Question: How many polls have there been in the history of the country? Maybe we can even narrow it down and ask "How many political polls are taken in a year"?

With 1000 respondents the sampling error is ± 3%, right?

That means that 68% of the time, the population will vote within ± 3% of the numbers the polls find.

95% of the time, the population will vote within ±6%

About 1 in a million of those polls will miss by close to 20%.

Think there have been a million polls yet?


Now, if they are consistently off, then yeah, they probably have a systematic flaw in their sampling or their weighting. But with so many polls conducted so often in the 24 hour news cycle world we live in, sampling error can have significant effects.

What Ike said .
:confused:

SanJoaquinSooner
2/6/2008, 05:23 PM
Ummmm....yes it can.



With 1000 respondents the sampling error is ± 3%, right?

That means that 68% of the time, the population will vote within ± 3% of the numbers the polls find.

95% of the time, the population will vote within ±6%



.

They reported a margin of error of ± 3.3%. Ordinarily, polls use 95% confidence for their margins of error, not 68%. Do you believe Zogby reports a margin of error with only 68% confidence?


....or maybe you're thinking about standard error instead of margin or error.

SanJoaquinSooner
2/6/2008, 08:23 PM
I might also point out, in its final poll, Zogby had Romney winning California by 7%, although McCain won by 8%.

Zogby had Obama winning New Hampshire by 13% in it's final NH poll, although Clinton won by 2.6%



Originally Posted by Ike
Ummmm....yes it can

and I could win the superlotto three weeks in a row.