Well, then we're getting off track if your main point is about how the sampling is done, because we don't know in this case (yet). However, this is hardly a novel problem, and there is no particular reason to think that this study *will* have major sampling problems.
But if they do, doubling or tripling the number of subjects is not going to help.
Given the vast number of cultures capable of being produced within the US, there is no assurance that 2000 can hit them all, and realize what they represent.
Depends on what you want. Just because this sample may not sample any 2nd-wives-of-interracial-polygamous-marriages doesn't invalidate the apparent aims of the study. I don't think the researchers wanted to sample the full, intricate detail of all possible variation in attitudes (unlike Kinsey, say, who really was interested in specific variation as much or more than simple averages).
My argument is that the number 2000 given the demographics of the US and what they are looking for prevents A from being possible (or I should say, accurate).
Then we're speaking different languages. By what I mean by point A, if the sample size was 3 or 3000 or 3,000,000, A could be true.
" trying to show you need to take into account physical demographics in order to properly get a random population sample. "
If we had an accurate list of the entire U.S. population, we could sample randomly and never have to think about physical demographics one bit. As it is, we consider these sorts of issues only as a remedy for known biases in our sampling procedure.
"I'm sorry did I say that was an example of absolutely definitive research? "
Huh? I'm sorry, but did *I* say that *you* said it was "absolutely definitive"???? I was very careful not to put words in your mouth and very purposefully used your own terminology: "useful". You're the only person to use the phrase "absolutely definitive". Until me, right now, twice in this paragraph, but that doesn't count.
Why does that mean I am thinking that we'd need some outrageous number?
Don't get your knickers in a twist. There was no intent to say that you said that. I'm saying he had a similar number of subjects. "Millions" was an extreme contrast. I just don't see why Kinsey's same-order-of-magnitude-sample-size-with-highly-biased-sampling was being contrasted at all with the OP study.
There were perfectly valid reasons for Kinsey to get a few thousand more subjects, IMO. Doesn't mean that all studies need it, or that 2000 is necessarily insufficient.
Honestly, are you suggesting that a researcher would actually look down on a study that involved 10K as opposed to 2K samples?
Well, yes, I would in many cases. For many if not most purposes, 10K would be an enormous waste of money and resources. It would indicate that the folks running the study had too much money and too little knowledge of statistics. They could have run 2, 3, 4 or more studies for the same amount of money and obtained 2, 3 or 4 times the useful information.
As for voting vs. opinion polls:
Well, N=2000 wouldn't give enough accuracy for an election. +/-3% ain't gonna cut it. And the problems with proper sampling change from usually small but irritating biases to changing the course of history. That said, I think one could actually make a pretty darn good case for the superiority of random sample polls vs. voluntary elections, if your intent is to elect leaders who conform to the preferences of the largest number of people.
Elections don't work like that. They're about who gets the most votes, which is in principal similar but in practice not the same thing.
™ Version 4.2