PPP/DailyKos weekly poll has Obama up 8

[Alcon]

No one promised anything. And sure, PPP was only off by 4 pts, but that's pretty poor compared to how many pollsters GOT IT RIGHT, and pretty poor for a pollster everyone claims is the best

adsfasf.

This is not how statistics work!

An absolutely perfect pollster will FREQUENTLY have nights like this, even if they are doing everything perfectly with a perfect sample.

I'm not saying that there isn't some aggregate bias in PPP, but this is an absolute failure of an argument demonstrating any such bias.  Have you even calculated the probability of this size of a miss with a perfect sample?  Have you estimated a house effect using previous polling results?  Anything?

Or are we seriously going, "Idk about statistics, but other pollsters got it right"?  Oy vey.

[Alcon]

"Bias" in statistics is not the same as "house effect," which matches the colloquial definition of "bias."  Statistical bias is any deviation from the population parameter.

I read exactly what you were doing.  "But that's pretty poor compared to how many pollsters GOT IT RIGHT" -- Why do you find this meaningful?  You understand that pollsters cannot possibly do better than a perfect sample?  If a pollster averages out to hit within MoE more often than 95% of the time, that does not mean they are better.  That means they are lucky.  Accordingly, observing that other pollsters did better than PPP in one race is a fairly useless analysis.

I replied to your post after reading this comment in another thread: "After a stellar polling job in last night's Wisconsin recall (NOT)."  This comment is misleading.  PPP could have done a perfect job of polling, and a perfect job of predicting turnout, and ~20-25% of the time (just guesstimating) they still would have missed by about 4 percentage points.  You have absolutely no idea from your analysis if they didn't do a "stellar" job of polling.

[Alcon]

"But that's pretty poor compared to how many pollsters GOT IT RIGHT" -- Why do you find this meaningful?  

I'll send you the updated Nate Silver rankings when they come out. That's how it is meaningful.

You can defend that statement, I suppose, (although it will have a very marginal effect) but how do you defend "After a stellar polling job in last night's Wisconsin recall (NOT)"?  You have rather low confidence (probably less than ~70-75%) that's true.

"Bias" in statistics is not the same as "house effect," which matches the colloquial definition of "bias."  Statistical bias is any deviation from the population parameter.

Repeatedly missing in 1 direction like they did with Amendment 1 is a sign of putting out nonsense.

1. The average systemic bias on gay marriage polls (-7) is greater than the average systemic bias on other polls, indicating it may not just be the pollster's screw-up.

2. Show me something with actual statistical analysis and I will entertain your conclusions.  Otherwise, how do you not know you're accidentally cherry-picking?

[Alcon]

Compared to the polls released in the last 2 weeks, PPP did not have a good performance. About 5 pollsters were more on the mark than them in the Wisconsin recall. That's how statiscians, like Nate Silver, rank pollsters. That's also why I said "After a stellar polling job in last night's Wisconsin recall (NOT)." I said that because, comparatively, PPP missed the mark more than about FIVE OTHER POLLSTERS. You say that doesn't matter. Then send Nate Silver an email.

I did not say "that doesn't matter."  You are consistently imprecise in both your debates and numerical analysis.

[Alcon]

I did not say "that doesn't matter."  You are consistently imprecise in both your debates and numerical analysis.

Right. Thank you, godfather.

I don't know what that's supposed to mean, but whenever you'd like to respond to points actually included in the words I typed...