Another pathetic NPR poll

[hoosierdoc]

NPR: 8-point swing puts Romney in front | WashingtonExaminer.com

So... they interviewed 1000 people nationwide, or 20 per state

They polled heavier in 12 "battleground" states and hit 462 people there. So they interviewed 38.5 in each state and they feel confident in their numbers? They said a 3% margin of error, yet each respondent in a battleground state represented almost 3% of the response in each state.

I fail to see how this poll is worth anything, but it did show Romney ahead and up 8 points from a month ago

 

[Birds Away]

You distrust National Progressive Radio? I am shocked!

 

[hoosierdoc]

You distrust National Progressive Radio? I am shocked!

Generally yes, but the sample size seems woefully inadequate on this one. Why is the government sponsored agency paying for polls?

 

[Willie]

"What's more, those polled said that Obama, by a 55 percent to 44 percent margin, has spelled out a clear agenda for the nation." - shows how skewed it is......

 

[neraph]

NPR: 8-point swing puts Romney in front | WashingtonExaminer.com

So... they interviewed 1000 people nationwide, or 20 per state

They polled heavier in 12 "battleground" states and hit 462 people there. So they interviewed 38.5 in each state and they feel confident in their numbers? They said a 3% margin of error, yet each respondent in a battleground state represented almost 3% of the response in each state.

I fail to see how this poll is worth anything, but it did show Romney ahead and up 8 points from a month ago

Yes, a sample size of 1000 will have a approximately 3% margin of error for a 95% confidence interval. Which is to say, if they conducted this poll 20 times, 19 out of 20 times the results would be within 3% of each other.

Sample sizes for most polls fall between 1000 and 3000.

Check out https://en.wikipedia.org/wiki/Margin_of_error
and https://en.wikipedia.org/wiki/Statistical_sampling

Yes, statistics are hard.

 

[hoosierdoc]

Yes, a sample size of 1000 will have a approximately 3% margin of error for a 95% confidence interval. Which is to say, if they conducted this poll 20 times, 19 out of 20 times the results would be within 3% of each other.

Sample sizes for most polls fall between 1000 and 3000.

Check out https://en.wikipedia.org/wiki/Margin_of_error
and https://en.wikipedia.org/wiki/Statistical_sampling

Yes, statistics are hard.

BUT... you're looking at the entire poll +/- 3% but they are still claiming that margin on their smaller sample size of 462 for the battleground states. Really it is a whole bunch of little polls all combined into a meta-poll, right?