Shorter Nate Silver:

I'll fit a regression model with 8 variables, all of which are statistically significant. But when I want to do my analysis I'll use a reduced form and count only 2 of the 8 variables and pronounce Franken the winner.

Of course he could still turn out to be right, unfortunately his logic and model are wrong.

To use the model to predict the winner you'd have to have the total number of challenges initiated by each campaign and then do the crunching. There is no justification for tossing the variables he's included simply because in December there wont be anymore challenges. That is true, but we don't know how many there will be in each precinct so his projection/prediction is based on faulty logic.

My contention is that the perception of legitimacy is at least as important as accuracy in elections decided by a fraction of a percent of the vote.

On that, I whole heartedly agree with you.

lacking any evidence to the contrary (your vague misgivings do not count, James), I suspect the recount process is fair

The problem is that there's no way to know. Even those of us who follow American politics fervently basically have to take their word for it. When the process consistently overturns Election Night result, it's simply natural to question the legitimacy of the process. My contention is that the perception of legitimacy is at least as important as accuracy in elections decided by a fraction of a percent of the vote.

My serious question would be to ask if you actually have a significant sample size?

Statewide elections that are so close as to require a recount are rare. But we have the 2000 presidential election, the 2004 Washington governor's race, the 2002? North Dakota Senate contest, the 2008 Alaska Senate race, and the 2008 Minnesota Senate race off the top of my head.

My serious question would be to ask if you actually have a significant sample size?

Unfortunately, that leads to a human decision having to be made as to the voter's intent. And those humans have preferred outcomes. It's a process fraught with peril.

Seeing as both candidates have representatives there to challenge any ballot where intent is unclear, and that those humans have preferred competing outcomes... and lacking any evidence to the contrary (your vague misgivings do not count, James), I suspect the recount process is fair,

Leaving the initial results stand unless a strong case is made that fraud occurred?

In this day of electronic voting, it is even more important, especially in close elections. We have the scan machines here in Crawford Co and if it were not for that paper trail, I would have had no faith in the system. We have been using these machines for several election cycles, and each time the pens have changed. Why? I suspect that they do an audit of the votes after the elctions, find problems, and then try to correct them, as a good non-partisan election board should.

Fraud is the least of the problems that can occur. For instance, why did it take so long to declare MO for McCain? The rest of the country was weeks ahead of us. I would really like to know the answer to that question.

Didn't you just mention Florida 2000?

It ultimately didn't go to statewide recount. I have little doubt that Gore would have prevailed had it done so.

But what is the alternative to a hand recount? And if there is none, what can be done to attenuate the peril?

Leaving the initial results stand unless a strong case is made that fraud occurred? And then counting only ballots in which the instructions were followed?

And, of course, making ballots uniform and simple to begin with.

"I can’t recall a statewide election that ended closely and required either a recount or going to the absentee ballots in which the Democrat didn’t come out ahead."

Huh? hello....
Didn't you just mention Florida 2000?

It's a process fraught with peril.

There is truth to that, James (but I suppose one could say that democracy itself is a process fraught with peril--depending on your view of an election's outcome). But what is the alternative to a hand recount? And if there is none, what can be done to attenuate the peril?

'Tis to laugh. But what else do the Bush dead-enders have to talk about?