At 11:28 PM 11/1/2004 -0600 Julia Thompson wrote:
>> > > I know many of you see the same articles on ABCnews.com that I do,
>> but I ran
>> > > across this website buried in a technology article, so some of you
>> may have
>> > > missed it.
>> > >
>> > > This website shows the predicted electoral college votes per state
>> based on
>> > > polls.
>> > >
>> > > Kind of interesting...
>> > >
>> > > By the way, this site is taking 600,000 hits a day so it may take
>> a few
>> > > seconds to load the main graphic of the US map...
>> > >
>> > >
>> > > www.electoral-vote.com
>> >
>> > Oh, you mean the site that John and I discussed last week, John
>> slamming
>> > it for using polls that don't have very good accuracy? :)
>> >
>> Unfortunately for John, this site is the most respected electoral site
>> on the web, getting 6 times as many hits as its closest competitor.
>> Try reading this as it is very very interesting:
>> http://www.electoral-vote.com/info/votemaster-faq.html
>>
>> xponent
>> Visits Most Electoral Site This Cycle Maru
>> rob
>
>Yes, it's the most popular site. It's spending a lot on advertising to
>achieve that status, as well, at least that's the impression I get from
>the various postings.
>
>But John was arguing with the use of certain polls there.
Thanks Julia.
I mean, who cares if it is the most visited site? And since when does
visits equal "respect"? Would the opinion of political professionals and
academic pollsters count for something in the "respect" department?
Internet insta-polls get tons of hits too, does that mean that they matter?
And isn't USA Today the nation's most read newspaper?
And indeed, this is primarily a matter of statistics - and I would hardly
expect the most accurate statistical analysis to be determined by a hit
counter.....
Anyhow, as Julia noted, a huge problem with this site is that they include
Zogby polls. Zogby uses an unproven Internet-based methodology (who knows
who many 10-year-olds he has reporting) and weighting procedures that are
both opaque, and are considered unreliable by academic pollsters.
Likewise, it gives equal weighting to unproven robot-calling pollsters
(again, who knows how many 10-year-olds are answering) as it does to more
trusted pollsters.
Furthermore, I object to this site because it uses a "last poll"
methodology. This is ridiculous for a number of reasons. First of all,
we now have several polls being released per day - to arbitrarily select
one as the "most recent" and to ignore the others is simply ridiculous.
Secondly, it is simply bad statistics. If the true population is Bush 48
and Kerry 47, you are more likely to "hit" the true population distribution
through analyzing a number of random samples in succession than you to do
so by looking at a single sample from the population. Lastly, this site
has consistently produced wacky results which we simply know to be not
true. At one point this year, they had Utah going for Kerry - only the
single most pro-Bush State in the Union. They also had Colorado going
for Kerry even as Kerry was pulling ads out of the State and cancelling
visits.
The following site, while it isn't perfect, uses a much sounder methodology:
http://www.uselectionatlas.org/USPRESIDENT/GENERAL/CAMPAIGN/2004/polls.php
JDG
_______________________________________________
http://www.mccmedia.com/mailman/listinfo/brin-l
