M
Mark
Guest
A Muzi wrote:
>>> "Edward Dolan" <[email protected]> wrote:
> -snip the usual-
>
>> "rotten" <[email protected]> wrote in message
>>> Why did the single payer referendums fail in Oregon and Massachusetts
>>> then? The fact is that while people acknowledge there are large
>>> problems with our health care system, if you look at polls you'll find
>>> that people are satisfied with their own personal healthcare.
>
> Amy Blankenship wrote:
>> How often do polls reach people without phones?
>
> Good point but statisticians have largely corrected for that, noting a
> margin of error which includes both that and other anomalies.
Well, actually, the reported margins of error routinely line up with
those for 95% confidence intervals for simple random samples of the same
size.
In non-technical language, if they /hadn't/ taken phoneless people into
account, and used the most common margin-of-error estimate of random
sampling error, in most cases they'd get the _same_ margin of error as
is reported in the press.
I check this a lot, since I routinely use survey reporting in teaching
my statistics classes.
It's not that the pollsters aren't smart, they certainly know that the
phoneless (and the land-line-less) could be a cause of systematic error
in their polls. Further, most(!) pollsters have strong incentives to
accuracy (it's what they sell) - they don't want another embarrassment
like calling the '48 race for Dewey over Truman.
On the other hand, with the exception of election predictions, there are
very few opportunities for phone-only polls to be proven wrong because
of missing the phoneless (they will only be compared to other phone-only
polls, seemingly), and you can even argue away weak election predictions.
It's possible that the pros (Gallup, etc) have made estimates that show
that !currently! the land-line-less aren't systematically different
enough (or numerous enough) to make a difference. Or it could be that
the problem is being brushed under the rug because there's little to be
done about it.
Mark J.
You'd
> have to imply that unlisted persons as a group are different from listed
> persons as a group in a significant way to worry about it.
>
>>> "Edward Dolan" <[email protected]> wrote:
> -snip the usual-
>
>> "rotten" <[email protected]> wrote in message
>>> Why did the single payer referendums fail in Oregon and Massachusetts
>>> then? The fact is that while people acknowledge there are large
>>> problems with our health care system, if you look at polls you'll find
>>> that people are satisfied with their own personal healthcare.
>
> Amy Blankenship wrote:
>> How often do polls reach people without phones?
>
> Good point but statisticians have largely corrected for that, noting a
> margin of error which includes both that and other anomalies.
Well, actually, the reported margins of error routinely line up with
those for 95% confidence intervals for simple random samples of the same
size.
In non-technical language, if they /hadn't/ taken phoneless people into
account, and used the most common margin-of-error estimate of random
sampling error, in most cases they'd get the _same_ margin of error as
is reported in the press.
I check this a lot, since I routinely use survey reporting in teaching
my statistics classes.
It's not that the pollsters aren't smart, they certainly know that the
phoneless (and the land-line-less) could be a cause of systematic error
in their polls. Further, most(!) pollsters have strong incentives to
accuracy (it's what they sell) - they don't want another embarrassment
like calling the '48 race for Dewey over Truman.
On the other hand, with the exception of election predictions, there are
very few opportunities for phone-only polls to be proven wrong because
of missing the phoneless (they will only be compared to other phone-only
polls, seemingly), and you can even argue away weak election predictions.
It's possible that the pros (Gallup, etc) have made estimates that show
that !currently! the land-line-less aren't systematically different
enough (or numerous enough) to make a difference. Or it could be that
the problem is being brushed under the rug because there's little to be
done about it.
Mark J.
You'd
> have to imply that unlisted persons as a group are different from listed
> persons as a group in a significant way to worry about it.
>