Why?

[SO]

But the fact that you and other posters such as \"So\" will link a liberal site in here or another forum and say \"See I told you...\" (paraphrasing) is ridiculous.

Listen, you asshat. All I've said is that your statement that "polls aren't credible" is fallacious. So I linked to a graphic from a liberal website...who gives a fuck? Whether he's republican, democrat, commie, or nazi, it doesn't matter because facts are facts period. Did he fabricate the raw data that he used? My reason for posting it, if you'll go back and look, was to illustrate that if you look at the results of numerous polls on the same topic, the results will be similar nearly every single fucking time. Why? Simple binomial statistics.

Standard deviation of sample proportion (p) = square root {(P*(1-P)/n}

where n = sample size.

In most cases, the sample standard deviation is about 1.5%

A 95% confidence interval for the sample (p) is:

(p) plus or minus 1.96*(Standard deviation)

or roughly 3%.

At a 95% confidence interval, if you took 100 random samples of this same size, that 95% of them would report a (p) within this interval.


Virtually all public opinion polls use a sample size of around a thousand, there isn't much point to sampling more. There is a diminishing returns aspect to sampling. If you sampled two thousand then you could say the "margin of error" was about 2% instead of 3%. So it costs you twice as much but the gain in terms of information is low.

Math is neither liberal nor conservative. It simply is, so stop looking for boogeymen where they don't exist.