In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] wrote:
<Snip>
> >
> > "For results based on the total sample of likely voters, one can say
> > with 95% confidence that the margin of sampling error is +/- 2
> > percentage points."
>
> Those guys are supposed to be professionals, and they should have
> been polishing their syntax for 50 years, but my first reaction to
> that statement was "UGH."
>
> Is that what you are pointing to?
>
> I think that I want to say, with 100% confidence, that I have computed
> certain results. Beyond that, I am not sure how to describe them to
> the public.
I was indeed pointing to that error in describing confidence limits.
The Gallup press release repeated the error twice:
"...one can say with 95% confidence that the margin of sampling error is
+/- 2 percentage points."
Here are some better phrases:
Our estimate of a candidate's percentage of the vote should be within 2
percentage points of the real percentages in 19 polls out of 20.
(Now, the astute reader might ask "What does this have to do with the
price of bread? You didn't take 20 polls did you?" The answer would be
no. But the pollster might say, that in an ideal world, the estimate
would be within 2% of the true value 19 times out of 20. Then, the
reader might ask whether Gallup had ever compared their estimates with
the true percentages. Then the pollster would probably say, "Well, the
true estimates are different from our estimates plus the margin of error
far more often than 5%, but that is due to our polls being inaccurate,
not imprecise.")
Another possibility:
We are 95% confident that our polling percentages are no further than 2%
of the true percentages.
(I don't like this cop out of using confidence instead of chance or
probability. It almost makes you want to become a Bayesian to avoid
these problems.)
I am disturbed that few newscasters or pundits discuss the accuracy of
the estimates, just the precision. Freedman, Pisani, and Purvis's
"Statistics" has a very nice section describing the accuracy of the
Presidental polls up to 1992. The polls are often far off the true
voter percentage, with the worst estimate in Freedman et al. being
Clinton's 1992 victory, where Gallup overestimated Clinton's winning
percentage by over 5% (I don't have the book in front of me). That is
far more than the margin of error. The bias was caused by a formula
that assigned too high a percentage of undecides to Clinton in the
pre-election polls.
The best article I've seen on the bias in this year's polls is Saletan's
"Pollish sausage" in Slate, whose link was posted earlier on this user's
group.
http://slate.msn.com/framegame/entries/00-10-26_92147.asp
After reading Saletan, I have as much confidence in the Las Vegas
bookies as in some of these polls. I was particularly disturbed to see
Saletan's argument that the Gallup organization may design its
survey to produce wild swings in the percentages, since changing poll
numbers sell papers and lead the news, "It's also practical. CNN and USA
Today are in the news business. They're paying Gallup for new numbers
every day. If Gallup's numbers don't change, where's the news? So Gallup
has an incentive to keep its filter loose, allowing the winds of
shifting partisan intensity to blow its numbers back and forth."
Sent via Deja.com http://www.deja.com/
Before you buy.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
