Tim Erickson, Dec 10 2007
To elaborate on David Bee's post (Dec 9 2007) "No, We Don't Want To Kiss And Make Up," in which Neyman and Fischer duke it out: another Stats slugfest is a-brewin', and it may be worth visiting later in the course.
Many groups are concerned about the potential for miscounts in upcoming elections. They propose to "audit" election results by doing a hand count of a sample of the precincts involved. But how many precincts? Policy up to now has been to do so (if at all) for a fixed percentage of the precincts. Well, any good AP stats student knows that a fixed percentage of the population is not the relevant figure, but rather a number sufficient to detect the smallest feared miscount. That depends on how confident we want to be and the apparent margin of victory—and that is all about power.
New Jersey and Utah are both in the throes of trying to set better policy, and what they do may affect other states. You can read interesting papers describing the issue from Verified Voting. "SAFE" stands for Statistically Accurate, Fair, and Efficient.
SAFE article from Verified Voting
.
These articles make interesting and timely points about the use of statistics in modern life and the concept of Power.
Controversy, Politics, and the Importance of Statistical Literacy
So has the statistics community come to the rescue of America, showing us the True Way? Well, no, not yet. Although we all agree that we need to protect our system from screwups or malfeasance, we don't all agree on whether the procedures will work.
Kathy Dopp, in particular, has been a vocal critic of the Verified Voting scheme.
Here is her critique of the "SAFE" paper. What
do you think? Is she an irritating loony,
or are her good ideas being willfully overlooked by an
establishment that won't take the time to listen carefully
to a hardworking mathematician who doesn't happen to
hold a Ph.D. (and, dare we add, is a woman)?
See http://electionarchive.org/ for more of her work.
What makes this so interesting and treacherous is that it is a collision of stats and politics. If you are at all worried about the integrity of our voting systems, some kind of audit makes sense.
But what happens in some legislative committee if one group of statisticians recommends one policy and another says, "no! That's mathematically unsound!" Will the legislators be able to judge whose plan is correct? Seems to me that the status quo will have the upper hand: we have to get our story straight if we want to make any change.
But we have one chance to get it right. Imagine the reaction if we put something in place and then in 2015 we have to say that our audit policy was insufficient. The public and State legislatures will think, "pointy-headed statisticians don't know what they want, we might as well use touch screens."
Whatever happens, I think it will be an interesting story—a triumph, an object lesson, or an ongoing struggle—for stats students and teachers of all stripes to weigh in on. The best path is annoyingly hard to see.