Re: Such a seemingly high figure only represents 5% of the population. That's the point. Archived Message
Posted by Der on July 28, 2022, 6:36 pm, in reply to "Re: Such a seemingly high figure only represents 5% of the population. That's the point."
Ian: "So why is it that the majority of qualified statisticians believe that Survation, YouGov et.al. polls with a sample size of about 2000 or so people from the UK population of about 70,000000 is representative enough to reliably obtain an accuracy of about +/- 3% ?" Obviously, their sample is random. Imagine if they were confined to the same sample, i.e. the exact same people, every time they did a poll. Polling would be worthlesss. Kagarlitsky says only 5% of the Russian people are interested in doing polls. Presumably, the exact same 5% every time - some people are into polls some are not (he talks more about this in the earlier post by Gary). Thus, not random, and thus not representative. Hence worthless. And YouGov et al. certainly wouldn't get away with talking about a margin of error of +/-3% if their sample wasn't random or if it was biased in some other way. Ian: "Do know something about this highly complex subject that they don't?" Getting snidely are we? Well it does sound as if you might learn a thing or two from David Spiegelhalter's recent book "The Art of Statistics - Learning from Data". Unfortunately he doesn't cover bad manners so you'll have to look elsewhere for a cure for that. I thought you were Ian that's the only reason I responded. I'll just ignore you in future. I believe we're done here.
|
|