Quote:
Originally Posted by Slava
Yeah, but it’s voting distributions that matter. These polls that have say 2000 people surveyed in Alberta and try to predict an election outcome are really going to struggle. I’m not sure that 2000 Calgarians is a good sample to try to get to seat counts.
|
I think with a 2000 person Calgary sample (assuming that polls actually give you a representative sample and not a biased sample) can give you good riding level info by applying the partisan lean of a riding relative to the total of Calgary. Reasonable sample sizes from individual ridings are better but the risk of a non-representative sample goes up as sample size goes down.