Boffoblog: No Confidence in Berkeley Study

March 2005

November 20, 2004

Sun	Mon	Tue	Wed	Thu	Fri	Sat
		1	2	3	4	5
6	7	8	9	10	11	12
13	14	15	16	17	18	19
20	21	22	23	24	25	26
27	28	29	30	31

No Confidence in Berkeley Study

This will be a brief post on the debate over the "missing" 130,000+ votes in Florida. There will be more to come from me before the weekend is done, in particular an attempt to sort out the competing claims from the gearheads, I hope in language that the non-gearheads appreciate.

The much-discussed Berkeley study which reaches this conclusion has some methodological problems which ought to make us quite cautious. For now, I will make only one simple point. Take a look at Tables 2-6 and then look at Figure 1. See anything missing from the graph? Here's a hint: These are estimates and not certainties.

As Stats 101 will tell you, an estimate contains error. A poll, for instance, will have a margin of error of, say, plus or minus about 3% with a sample size of 1,000. That is, with 95% certainty, the population proportion will be within 3% above or 3% below the sample's proportion.

So, what is the 95% confidence interval for the point prediction the Berkeley team makes that 130,733 votes "disappeared" from Kerry's column? They don't tell us. In other words, we do not know whether, at any reasonable level of confidence, the electronic and optical scan methods really did produce substantively meaningful differences.

Moreover, when graphing the estimated effects as they do in Figure 1, the confidence interval will be wider -- that is to say, more uncertain -- the closer one gets to the tails of the distribution (one end of the sample or the other). Recall that nearly all of the power of their conclusion comes from the three large Dem-leaning counties on the far right of the graph. In other words, the counties that are closer to the fat part of the (for now invisible) confidence interval. Indeed, except for that end of the distribution, the two voting methods track very closely. This makes their lack of attention to the confidence interval all the more troubling.

At a minimum, this is a shoddy way to report statistical results. Worse, the interpretation may be downright misleading without taking into account the confidence interval. They are advertising an estimate as certainty. It's just a matter of basic statistics. Let's not overhype it until they show their work.

01:39 AM in Politics | Permalink

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d83421b9a553ef00d83421bb7153ef

Listed below are links to weblogs that reference No Confidence in Berkeley Study:

» Berkeley researchers: "Irregularities May Have Awarded 130,000 - 260,000 or More Excess Votes to Bush" from The Command Post - 2004 US Presidential Election
From their press release: Today the University of California’s Berkeley Quantitative Methods Research Team released a statistical study - the sole method available to monitor the accuracy of e-voting - reporting irregularities associated with ele... [Read More]

» differing thermohaline from differing thermohaline
york feedback upper technology trends orbital [Read More]

Verify your Comment

Previewing your Comment

Posted by: |

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:

Your comment has been posted. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.

Boffoblog

substandard deviations

Recent Posts

Categories

March 2005

Archives

I Recommend:

Bottom of the Page

November 20, 2004

No Confidence in Berkeley Study

TrackBack

Comments

Verify your Comment

Previewing your Comment

Post a comment

politics

divertissements

ruminators

sound & vision