One of the virtues of TV …

November 4, 2008

… is that it doesn’t go down when you try to refresh the site to get the latest election results.  Not like some sites I know … <cough>, <cough>CNN<cough>,<cough>.


CNN – dropping the ball on stats is our game!

September 15, 2008

Bonus question from today’s coverage on CNN of the American presidential candidates’ responses to the melt-down on Wall Street today:

If economic problems continue to dominate the headlines, they could help Obama in the tight race for the White House, the recent CNN/Opinion Research poll suggested.

According to the poll, Obama was viewed as being better at handling economic issues by 52 percent of the voters surveyed. In comparison, McCain was viewed as better on economic issues by 44 percent. The margin of error on that question was plus or minus 4.5 percentage points.

Take a good look, and decide for yourself what the problem is with the conclusion from those two paragraphs of text.

[ … ]

[ … ]

[ … ]

[ … ]

No dice?  Here, I’ll play it again, with some boldface on the font to help out:

If economic problems continue to dominate the headlines, they could help Obama in the tight race for the White House, the recent CNN/Opinion Research poll suggested.

According to the poll, Obama was viewed as being better at handling economic issues by 52 percent of the voters surveyed. In comparison, McCain was viewed as better on economic issues by 44 percent. The margin of error on that question was plus or minus 4.5 percentage points.

Got it yet?  Let’s assume the percentage plus or minus the margin of error is the 95% confidence interval on the true proportion of Americans who would rate Obama or McCain as being better on economic issues.  The 95% confidence interval is the range of values within which we are 95% sure that true population value lies, and to get that range, you take the state proportion and add or subtract the margin of error.  For Obama, that would 52 +/- 4.5 = {47.5,56.5}.  For McCain, that would be 44 +/- 4.5 = {39.5,48.5}.  I could blather on about this for a while, but the short story is this:  based on their own numbers, CNN is saying that from the total American population, it could easily be the case that 47.5% think Obama is better on the economy and 48.5% think that McCain is better (the remaining 4% would be “neither” or what have you).  In other words, it’s a distinct possibility that Americans think that McCain is better than Obama on the economy, not the other way around! This means that the poll is not supporting the conclusion drawn in the previous paragraph, and someone around there should get a whack upside the head.

(Note: if you assume that there are only two poll choices – “Obama” or “McCain” but not “neither” or “no response”, or what have you – it can be more informative to look at the difference between the percentages and see if that is statistically greater than 0.  A few numbers on the back of an envelope is more than enough to show that p >> 0.25 here.)