MS-Change Research: Hyde-Smith +5

[cinyc]

Can someone explain to me how a poll of 1300 people randomly distributed across Mississippi can be junk?

Doesn't statistical theory say that is more than enough of a sample size to be accurate?

You could poll 10,000 Mississippians. If your methodology stinks or weighting is off, it won't matter one bit that the sample size is larger.

A one-question Google Survey with a sample size of 10,000, for example, would still be crap because of bad methodology.