Re: confederacy of dunces*

Date view Thread view Subject view Author view

From: Jeff Bone (
Date: Tue Nov 28 2000 - 19:35:41 PST

> That data doesn't say anything about the margin of error for the difference
> between a given machine count and hand count. We would need a third number -
> the actual vote count - to know the margin of error for the difference. You
> could also use the margin of error for each machine, versus the margin of error
> for any hand count.

This is all absolutely true. The point I was trying to get at, though, Roy's nice
data notwithstanding, is that it's impossible to make assessments about the margins
of error involved --- human or machine --- given just this data. You have to assume
one or the other is constant across like instruments in order to measure the other.
This isn't a safe assumption; apples, oranges, and all that. Either claim could be
true, given either contraverse assumption: either the machines could have wildly
dissimilar margins of error, or the humans could. No way to know which is the case,
given the data, but my assertion is that the latter is naturally more suspect.
(Common sense.)

> But the margin of error of a hand count doesn't have to be a constant to support
> my point. It only has to be smaller than the margin of error of a machine
> count.

True again, but there's no way to guage the former given the data without the
assumption of constancy on some other measurement --- which we can't do.

> If the machines are more likely to make errors,

We can't assume this.

> and there are errors,
> then errors are more likely to have come from the machines.

That's a tautology. So what?

> Thus a fluctuating
> error count is more likely to come from the machines than the humans.
> In any event, the voting machines downstate (Gore turf) were different than the
> ones upstate (Bush turf).

Still can't draw quantitative conclusions from that fact. Were the ones upstate
more "accurate?" That would be the only reason for selectively throwing out
measurements from the ones downstate, while retaining the upstate ones --- and then
only if you could prove that the handcount margin of error downstate was equivalent
to the machine margin of error upstate.

> ...good god these threads eat time. I am going back to lurking.

Heh. Pls. don't, you've actually been a voice of sanity among the rabid opposition
(ahem, Silly. ;-) More fun to hang in there. And eventually, these don't consume
that much time at all. After a while, you can easily FoRK 100 a day with almost no
effort. I'm actually working on some software to automate this process.



Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Nov 28 2000 - 19:41:04 PST