Ron Paul in statistical tie in California

[bjrnet]

According to the latest CNN/Los Angeles Times/Politico poll (released January 14) Ron Paul is in a statistical tie in California with McCain, Romney, Giuliani, Huckabee and Thompson. http://i.a.cnn.net/cnn/2008/images/01/14/relca1.pdf

Ron Paul at 8% !!!

Thompson is at 6%, Huckabee is at 13%, Giuliani is at 14%, Romney is at 16% and McCain is at 20%.

With a +/- 6 point sampling error he is statistically in a tie with all of the front runners.

[brandon]

lol, he is not really in a statistical tie. Lets take McCain as an example. There is only 1 way he could tie mccain. If mccain is unlucky enough to get the -6 spread, AND Paul is lucky enough to get the +6 spread. There are about 143 ways McCain could beat him, for example Paul gets +5 and McCain gets -6.

That means there is about a 0.7% chance of McCain and Paul tieing. I wouldn't call that a statistical tie. :) I like the optimism though

[pengieh]

I believe that the margin of error is using a 95% level of confidence, which would mean a ~0.25% chance that the data reflect an actual situation of Ron Paul at 14% and McCain at 14%.

[AlexMerced]

the best thing about california, even with 8% longs as they are spread across the rights area, we cans till get more delegates than the other candidates

[bjrnet]

lol, he is not really in a statistical tie. Lets take McCain as an example. There is only 1 way he could tie mccain. If mccain is unlucky enough to get the -6 spread, AND Paul is lucky enough to get the +6 spread. ...

That is within the margin of error and thus, if I understand it correctly, considered a "statistical tie". Note that we are only talking about polling 255 random and potential Republican voters here and the margin of error is high.

If there was a greater number of polled voters performed, let's say 1000, the standard error would probably drop to 5% or less AND if the same results were found -- then McCain would have a statistical lead over Paul.

Not that any of this really matters. But it is fun to be optimistic!

[jasonuher]

Not that any of this really matters. But it is fun to be optimistic!

This is true, but remember that most people don't know what a 'stastical tie' is, in order to be a true statistical tie, the movement has to be within half of the margin of error.

For example, if I had 10% and you had 20%, a margin of error of 5% would not put us at a statistical tie, we would need 12.5% and 17.5% respectively. Technically it is possible that the top number will move down 5% and the bottom number will move up 5%, but the probability of this happing makes it anything but a tie.

Also, someone mentioned confidence intervals, they doesn't really affect the numbers themselves; a confidence interval just says the likelihood that the results are in fact +- the margin of error. So if we had a margin of error of 5% and a 90% confidence interval, that just means that if we took the survey 10 times, the numbers would be +-5% what were quoted in the results only 9 times.

The last thing to remember is that phone polling is total crap. It is nearly impossible to get a representative sample, nearly impossible to phrase questions in an un-ambiguous manner, and nearly impossible to get honest answers out of people.

In the end it's all about which questions you ask, and who you ask them of; which are easily manipulated to receive the desired outcome. You can play with margins of error and confidence intervals all day, but if you just ask the right question to the right people you can have any result you want.