and some polls are better than others.  Also, what you can take from a polls is often different depending on what you want to see.

However each poll has something to say and we are discussing this one today.

think is actually accurate?

LOL!

Okay...who's gonna reimburse me for the coffee I spit on my screen!

Just didn't see that one comin'...too funny.

Thanks for the chuckle.

Just what I was thinking. The Clinton spin was moving from "She's going to take PA in a blowout!!" to "Obama has to win PA to stay in the race!"

Meanwhile, another poll shows Obama still in double digits nationally among Dems, and still another has Clinton falling out of the margin of error behind McCain.

Those polls, of course, don't count.

http://tpmelectioncentral.talkingpointsm emo.com/polltracker/

So this poll hasn't hurt me one bit.

Real expectations: If Hillary can't win by 15% she's toast.

I think that SUSA and Quinnipiac have been the two most accurate pollsters this year.

Quinnipiac in fact has been pretty stellar on the dem. side of things (never off by more then 5 points). They were 3pts off in FL and NY. 5pts off in NY and OH.

SUSA has been hit or miss on the Dem side. They were off by more then 10 points in NY (10), AL (12), SC (16) and MO (12). On the flip, they got CA right and were 1 pt off in NJ and NY.

the Q has Clinton up 5. SUSA has her up 18. I expect that the truth lies in the middle of these two pollsters.

If Rasmussen is closer to Qunnipaic, I'm going to think that SUSA is an outlier. I usually trust SUSA but they too have been wrong a few times. Remember Missouri?  They had Hillary winning by like 6 and she lost.

Whose expectations? Pre-Super Tuesday, Obama's awesome predictions xls had him losing PA by 5 points.

http://www.politico.com/blogs/bensmith/0 208/Obamas_projections.html

I imagine its been updated, but other than overestimating Hillary in most contests, it's been pretty accurate.

I would be very surprised if the win was double-digits, but that's not as an Obama supporter. That's just someone who yet to see Hillary take a large state other than NY by anything approaching high double digits.

but they vote, I agree.  

Rally turnout may not necessarily translate into votes, but such turnout is a great sign.  Dean may not have gotten the nomination, but grass-roots activity that his campaign helped develop has contributed quite a bit to revitalizing the Democratic Party.  The same goes for Obama.  Whether he wins or not, we should be glad he has run.

If you remember correctly Hillary barely won NH, so are you insinuating that she will win PA by under 3%?

That was due to several single night Polls which picked up tracking on the weekend, but not on Monday... in PA were talking about Multi-Day larger sample polls.

I see no evidence by which to believe SUSA here.  Perhaps if we were seeing a steady down-turn in the national number (which were seeing the opposite).  An event to drive down Obama or up Clinton (which we have not seen in the past week).  Or, collapsing Obama numbers in other states.

Of course I have no idea if this SUSA poll fits into that 5% category.

That's according to their statistical model, most likely using the standard deviation derived from some sort of assumed Multinomial distribution.  However, polls in practice are outside the assumed MOE much, MUCH more frequently than the calculated stastical significance.

Why?  Stastical variance comes in two forms - process variance and parameter error.  The calculation of a confidence interval from an assumed distribution accounts for only process variance, or in this case, the likelihood that the percentages in your sample of people differ from the true percentages of the entire underlying population jsut due to random error.  However, and I don't know this for sure, I assume that the polling firms make no attempt to quantify the parameter error in their MOE.  Parameter error is also called model estimation error, and in this case represents variance in the final results due to incorrect voter turnout models or incorrect weighting of sub-strata of the sample data, or things of that nature.

Is this a significant source of total variance in political polling?  Yes, of course.  It is the reason actual results do not fall within the MOE according to the predicted stastical significance (95% or 97.5% or whatever is used).  Then why would a polling firm ignore the parameter error?  There are two reasons that I believe that polling firms ignore this, one practical and on mathematical.  The mathematical reason is that (unlike process variance) parameter error is impossible to calculate 100% accurately and very difficult to even estimate with any degree of confidence.  The practical reason is that these firms are also marketing firms.  Including an estimate of paramter error in your polls, from a marketing perspective, is like admitting straight off that your model is incorrect.  Not a great selling point, and after all, these people are a commercial venture, so marketing is a big part of what they do.

Incidentally, I also believe that this one of the major reasons that poll averages end up being far more accurate than individual polling firms, even when you take into account the reduction of process variance from aggregating polls.  In other words, the average of 5 unrelated polls on a race is more accurate than the aggregation of 5 Rasmussen polls.  Each pollster's proprietary model has intrinsic biases.  Aggregation of polls with many different models allows the biases of individual pollsters to cancel each other out.

What to make of all this for this race?  I don't know.  This seems like an outlier to me, but OTOH, S-USA has clearly established itself as the best pollster in the Democratic primary this season, so maybe they have hit on the correct model and eveyone else has it wrong.

It's not surprising that people on a sinking ship will grab at anything they think will float.

thanks, that is another bit of information I did not know about polling.  Can you explain in more detail?

Two news polls are up in a diary. They further indicate this one is an aberration.

Indystar.com

Politifact.com

Columbia Journalism Review

The Hill

Huffington Post

ABC News "The Blotter"

I went to Obama HQ in Scranton and asked for signs etc.. for some friends who are supporting him.  They didn't have any lawn sighs/  However they gave me some stickers with that image.
As a work of art it is great. As a campaign poster it works in some places better than others. It would not be my choice.

But the funniest part of the article is down farther were Olbermann tries to smear Rendell.  I loved the response from the Rendell guy...rolf, "there are better reasons than that to think he is the worst person in the world".  That cracked my up and it probably cracked Rendell up too.
Olbermann is making such an ass of himself.

Most likely, since they're in the general area of Casey influence. Also, isn't that where the recent Obama bus trip stopped?

Troll rating for speculating that Casey could be helping Obama with conservative Democrats, but hurting him among more liberal ones?

Whatever Kobi.

Has she said that Ted Kennedy is old and needs to step aside, called unions special interests, tipped off the Canadian embassy that anti-NAFTA statements made by her were merely "campaign stuff?" No.... that are the first term Senator from Illinois.