11/4 Amusing quote from http://democraticunderground.com:
"CNN's exit poll showed Kerry beating Bush among Ohio women by
53 percent to 47 percent. Kerry also defeated Bush among Ohio's
male voters 51 percent to 49 percent. Unless a third gender
voted in Ohio, Kerry took the state."
I found it amusing because, even assuming exit polls can't be
horribly manipulated, and aren't noisy, the conclusion still
doesn't follow (due to Simpson's paradox). Btw, I can't
believe I haven't found that site before, it's so fun to
read! -- ilyas
\_ It's fun to read because they ignore little things like the fact
that the exit polls only poll a very small number of places, not
every place and even if they did poll every place, they still don't
poll every person so it is not possible to predict an entire state
from a small number of precints like this. They also ignore the
possibility that Kerry supporters sought out exit pollsters while
Bush supporters ignored them. There are several other things they
ignore which shows a deeply fundmental failure to understand very
basic statistics and how exit polls are actually conducted. We
don't ask Zogby or exit pollers to pick our President. We vote to
determine that. And yes, democraticunderground is fun to read. :-)
\_ Could this explain the strong Kerry exit poll at first, and then
due to some events, the numbers started to turn around?
\_ This isn't anywhere near a simpson's paradox band.
\_ Please give an example (specific numbers) of how this is
possible.
\_ Sorry, I have to eat crow on this one, a third variable has
to be involved in the statement. -- ilyas
\_ Simpson's doesn't apply. Ilya is meta-trolling.
What is more interesting is that from Tuesday to now,
cnn's exit poll numbers for ohio have slid Bush's way.
\_ This is interesting, you're right. Last time a looked at
Ohio (and Iowa) exit poles, they were slightly in Kerry's
favor, now they are slightly in Bush's favor. What
happened? Late poling data?
\_ IT'S A FOCKING SCANDAL IS WHAT IT IS
\_ Hoo-kay.
\_ The question is, was Bush ahead in *every* exit poll where the
votes are hard to inspect (EV/no reporter access) or did these
authors just pick and choose the ones where the exit poll showed
Kerry ahead but he lost, ignoring the exit polls where it went the
other way and said "look the actual vote is X% ahead for Bush
compared to every exit poll, that's statistically impossible!"
I agree that the level of unbelievability is pretty high on this
one, but we are talking about Karl Rove, the CEO of Diebold and
an administration that continues to lie about saying X even when
confronted by direct video evidence of them saying X.
\_ DU is the left wing version of Little Green Footballs.
\_ The Union lost, GET OVER WITH IT ALREADY. |
http://csua.com/feed/