Quote:
Originally Posted by geeker2
|
You always look at the sampling methodology, and compare to the margin of error the poll publishes. Some pollsters have a great reputation for predictive accuracy, some do not. The contention made here, that you have to exactly match the percentage number per category within the population you are sampling, isn't true from a statistical accuracy sense. There are lots of other things that come into play. The PDF posted with this article doesn't even list the sampling technique for selecting who was talked to, that I can find.
I can get an accurate scientific study published using just 6 horses, if you treat the datum accurately and fairly in a statistical sense. And no, that is not "making numbers say what you want".
(sorry, had to repost this under me)