Actually, if they are chosen randomly, the margin of error in that poll is about 5.7%. So, not a great survey, but not all that nuts either.

In case anyone is interested in the "shortcut" method for determining the margin of error for a survey of any very large group (it actually may apply to small groups as well - I'm not a statistician, just someone forced to take far to many math classes), it is E = 1/sqrt(N), where E is margin of error in %, and N is the sample size. It's not precisely correct, but will put you within a tenth or a hundredth or something like that.

That also assumes, again, that the survey is completely random from within the group, which are means that it is distributed fairly evenly across subgroups as well. I don't know anything about this poll, but if it was done by a Japanese gaming magazine or something, it would be more accurate to say that XX.X% of Japanese gamers think Y, with 5.7% margin of error.

Also, of statistical interest is all those polls you see in magazines and newspapers that claim "XX% of Americans Y, margin of error 3%" (which is a very common margin of error for those sorts of polls) have only actually asked (roughly) 1000 people out of the 350 million in the country.

Click to expand...