Quote:
Originally Posted by Suave
I guess I jumped the gun a little there, I'm just too used to seeing people dismiss poll results because "only 1000" people were surveyed. And by "regular" polls I wasn't specifically referring to political polls but rather to the general polls we hear about everyday.
|
Well, I really only meant "only 1000" in a very strict sense--that the margin of error is a sizable smidge higher than it is for 2000--and that it leads to even smaller sample sizes at the regional level.
For me, if you're going to survey only 150 voters in the entire West, you might as well not bother. The margin of error is so high that the results become very dubious. This particular poll shows a Tory drop of 10%--but my feeling is that this is far likelier to be statistical noise, and with only 150 respondents that becomes even more probable.
Since prediction of electoral outcomes in Canada hinges so clearly on regional results, this is a big methodological flaw. But I wasn't questioning the overall methodology of using samples in general, in case that's what you thought.