Reply #131: Kelly: DO NOT INSULT your fellow DUers like that. [View All]

[redacted]

Do not insult the intelligence, knowledge and awareness of me and the many people here who agree with my point. You probably have no idea with whom you're speaking!

I've spent more time in university math, statistics, econometrics and computer science classes than YOU probably spent in all your time in college, period. And let me educate you about something.

Any time you get half a million people to agree about anything, which is what this poll shows, you don't need rigorous statistical sampling techniques to tell you with academic precision whether the results are generalizable to the population.

I find it almost comical when "some people say" (as the phrase gets used on Fox news) "Geez Louise, all but 10 percent agree in that poll. But know what? It means nothing because it's "NOT SCIENTIFIC." Like these people have any idea of what the science that makes good survey research really amounts to.

Well, guess what. Beyond a certain point, it doesn't need to be "scientific." The common sense of those interpreting the results is probably more important. You don't need to have a PH.D. in statistics to draw the conclusion from results this strong that a very significant portion of the population, if not a majority, agrees about a certain conclusion.

Keep in mind that these days, modeling techniques are so good that one probably doesn't need more than about 50,000 people to infer very accurately what 250 MILLION people are thinking at any one time. This is why the debate over the Census is so absurd. A census is way, way overkill. It's an anachronism.

In this case, we've got half a MILLION. You can account for a hell of a lot of sampling error when you've got half a million people participating in any poll. The smart money here will be on those who assert that there's substantial support among a large percentage of the population for impeachment. Just how large, we'll soon find out with more precise analysis and techniques. But anybody who says that the proportion can't possibly be a substantial one because the poll is "not scientific" has a LOT to learn.