Quote Originally Posted by Sam Odom View Post
we have baseball games going for GOD SAKES

BTW-- margin of error could be Mit by +9
If the margin of error is +/-3% then the radius of error is 6%, 6% up or 6% down, So if the reported figure is +5% then the 95% confidence range is +11% to -1%.

This is because polls don't just have one variable they have two. So say I take a poll, one candidate, candidate X, receives 50% and the other candidate, candidate Y, receives 40%, if the margin of error is +/-3% then the candidates true number with 95% confidence would be 53% to 47%, on the other hand the candidate who receives 40% could be 43% to 37%

So you would say, that poll says candidate X is winning by 10%, but he could be winning 53% to 37% or he could be winning 47% to 43%, so the most with 95% confidence you can say candidate X could be winning by is 16%, and the least is 4% +/-6% either way.

Similarly in our scenario Romney is winning by 5%, so 6% up, 11%, 6% down -1%, so within the true margin of error for a two candidate race, he could still statistically be losing with 95% confidence within the margin of error for a two candidate poll, so he is not winning outside what the margin of error means for statistic purposes.