[Yaelle]

Each time I read a poll, I wonder how the information is gathered. The first thing I think about is caller ID. Many people with full lives will check the ID and not take calls except from people they know. Many people deliberately avoid polls, and many people are simply too busy, even if they took the call, to continue with it. It sounds like the Zogby calls involved tens of questions (the site as quoted above says 96, which I am finding hard to believe).

Is there some kind of automatic screening going on, simply by who chooses to answer the phone and continues with a very long polling session? And how does that correspond to Actual American Voters, some of whom are busy and don't answer polls?

Also, the sampling strategy for lower population sizes: does this really work? Does it end up meaning that if only two Asian men were sampled, it would be extrapolated to signify some absurd percentage of Asian men prefer one candidate to the other?

I know there is some science behind statistics and surveys, and pollsters like to show how close they were AFTER an election, but I keep coming back to little things like Caller ID. Is anyone explaining that the samples are heavily weighed toward either people who do not yet have caller ID or toward people who love speaking with pollsters?

[Yaelle]

Ping to explain polling methodology!

[hflynn]

The methodology is simple. Build up the Republican that is most likely to lose if nominated for as long as possible. In the last few days before ballots are actually cast try to bring a little reality into the polls so that our reputation as pollsters can be salvaged.

[Yaelle]

Ping to help explain why we should believe polls, mathematically, please?

[Radix]

Zogby polls proven primary voters.

Everybody has an opinion, and lots vote in actual elections, but few vote in primaries.

Zogby targets actual past primary voters.

[Yaelle]

If we really “live and die” by the poll numbers around here, I’d like to know the basis of the science of calling and bugging people, and how it relates to real voters actually voting!

[Perdogg]

I think they gather the samples and then they get demographics and apply a statistic test which include error

error = sqrt ((p X (1-p))/(n-1))

p (candidate A) = 0.48
q (candidate B) = 0.52
n = 502

error = (sqrt(0.52 X 0.48)/(501)) =0.02

Candidate A = 48%
Candidate B = 52%
Margin of error = +/- 2%

If n > 33, then n-1 is not considered statistically significant.

[Checkers]

“Math ExpertsPlease explain”

Zogby, unlike say Rasmussen, just makes it up.

[trumandogz]

Many people with full lives will check the ID and not take calls except from people they know.

Even without talking to people with Caller ID one can impute what answers they would give.

[depressed in 06]

You have it backwards. An election that does not match a predetermined poll result is a rigged election. An electorate that goes against a poll result is too stupid to see that they have been deceived.

[Mobile Vulgus]

Remember that Zogby is a Muslim activist and anti-American. His polling is always suspect.

[jmaroneps37]

“Math ExpertsPlease explain..”
zombie is a democrat. ‘nough said?

[Petronski]

It would be easier to say “mix one cup mayonnaise in a bowl with half a cup of dill relish (drained), a tablespoon of worcestershire sauce, three tablespoons of lemon juice with pulp, and half a teaspoon of rosemary. Whip thoroughly and chill for two hours. Use within three days.”