PA: Temple University,-, Obama (D) 50%, McCain (R) 41% (SINGLE DIGITS)

[tokar]

http://www.realclearpolitics.com/articles/docs/TemplePollOctober27200.pdf

Temple?

Ya know, I dont mind university polls...I respect Suffolk, Quinnipiac, Muhlenberg, Franklin&Marshall; they have a history of polling.
But I can't stand these one-poll wonders...VCU, Univ of Wisconsin, Univ of Cincinnati...and now Temple.

[tokar]

http://www.realclearpolitics.com/articles/docs/TemplePollOctober27200.pdf

Temple?

Ya know, I dont mind university polls...I respect Suffolk, Quinnipiac, Muhlenberg, Franklin&Marshall; they have a history of polling.
But I can't stand these one-poll wonders...VCU, Univ of Wisconsin, Univ of Cincinnati...and now Temple.

Temple, as pointed out was the closest in 2006 on the Senate race; Cinci was fairly good in Ohio governor's race, IIRVV on both.  Muhlenburg was crap in that election.  The others are good, in certain areas.

Are you seriously considering a poll 2 weeks out (10/25/06) as a predictor of the race?

You really need to get your head on straight.  Taking a single poll 2 weeks out as a predictor is like taking a single run of Nate Silver's 10,000 simulations as a predictor of the race.

You never cease to amaze me.  Trash pollsters who have a deep track record of polling, but trust the pollsters who poll once an election and then disappear for 2 years.

[tokar]

http://www.realclearpolitics.com/articles/docs/TemplePollOctober27200.pdf

Temple?

Ya know, I dont mind university polls...I respect Suffolk, Quinnipiac, Muhlenberg, Franklin&Marshall; they have a history of polling.
But I can't stand these one-poll wonders...VCU, Univ of Wisconsin, Univ of Cincinnati...and now Temple.

Temple, as pointed out was the closest in 2006 on the Senate race; Cinci was fairly good in Ohio governor's race, IIRVV on both.  Muhlenburg was crap in that election.  The others are good, in certain areas.

Are you seriously considering a poll 2 weeks out (10/25/06) as a predictor of the race?

You really need to get your head on straight.  Taking a single poll 2 weeks out as a predictor is like taking a single run of Nate Silver's 10,000 simulations as a predictor of the race.

You never cease to amaze me.  Trash pollsters who have a deep track record of polling, but trust the pollsters who poll once an election and then disappear for 2 years.

Tokar, I'm saying that the one poll that got it, PA Senate, the closest to being right in the University Polls was Temple. 

http://en.wikipedia.org/wiki/Pennsylvania_Senate_election,_2006#Opinion_polling


Here are the results:

Temple Casey +16

Mulhenburg Casey +8

Quinnipiac Casey +10

Keystone +15

Rasmussen  +13

Actual: Casey +17.4

Fluke, maybe, but that is still the result.  I think Keystone in now F & M.

Im not denying that Temple had it at +16.  My point is that you shouldnt be using polls as predictors of the MARGIN...let alone from a pollster who polls once a month.
On another note, I guarantee you that if this were November 7th, 2006 and you were looking at those polls as they are listed you wouldnt be saying "oh casey has a cake walk".   You would be saying that Casey would probably do +5, +10 at most, but definitely not +17.4.

The last poll on USElectionAtlas was a +4 for Casey...on a side.


Polling is always all over the place, and Temple could have easily pulled out a +6 instead of a +16.  Its funny that you are praising Temple, who polled PA all of twice and pulled a +16 out of their ass, put ignore all the other pollsters who polled 10 times as much and continuously showed a strong victory for Casey.

[tokar]

I'm not "praising" Temple, merely stating fact.  I think it's a uni poll (and not even the most pro-McCain recent PA poll).

I really don't understand why you would look a USElectionAtlas "poll," for proof of anything. 

BTW, if you look at the poll that was the most off, it was Muhlenburg, the one you continuously cite.

Also, where do you get the idea that I'm "...using polls as predictors of the MARGIN...let alone from a pollster who polls once a month."  I'm commenting, accurately, on the polls success.

You are reading far too much into the line:  "I think they were close in 2006, but it's a uni poll."

Apparently, you missed the point after the comma.

Its more that you are looking too much into numbers and not into trends...