- A 36 year old worker at a sign manufacturer in Minneapolis, Minnesota returned to his employer's building shortly after losing his job and killed 4 people including himself (ABC News): http://abcnews.go.com/US/police-minn-office-shooter-kills/story?id=17344310#.UGWDMpj7KWg
- The man behind the anti-Muslim film "The Innocence of Muslims" was held without bond Thursday evening on charges that were not related to his film (CBS News): http://www.cbsnews.com/8301-201_162-57521939/man-behind-anti-muslim-film-ordered-jailed/?tag=AverageHero;leadHed
*** Be sure to vote in the weekly LME "50 days until the election" poll on the left side of this blog ***
Happy Friday! :)
ReplyDeleteSince the topic of polling methodology and accuracy has been popular lately, curious what those here and LME feel about the methods used at this site:
http://election.princeton.edu/about/
This website combines aggregate data from all available state polls conducted across all of the states, combines them into probability curves, and applies algorithms to account for different historical and demographic factors. There are a few papers with further details breaking down the analysis methods further, but that’s the general idea of it.
They’ve been around since about 2004, and using their modeling methods they predicted the exact electoral spread of the 2004 election and were within one electoral vote of the 2008 election. While a sample size of two isn’t anything to write home about, I think they’ve at least earned a level of credibility; particularly considering how difficult of a prediction that is to make with such accuracy.
My counter-argument would likely be that I’m sure there are many possible models that could predict the 2004 and 2008 elections accurately, and still show different results in this election, but at the same time I haven’t seen anything along those lines surface.
All of that said, if the methods here do indeed prove accurate, Mitt does have some catching up to do.
http://www.foxnews.com/opinion/2012/09/27/truth-about-2012-polls/
DeleteRKen - good morning and happy Friday!
DeleteI've read through the Princeton analysis, and it points to itself as a good standard for poll predictions. There is, however, one problem I've noticed: it says it relies heavily on polls and it doesn't exclude any. It then uses probabilities to calculate who will win that state. That's all fine, but it still does not address the unrealistic oversampling of democrats in that state. If the Meta-analysis uses all state polls, and each poll over samples democrats by +10, wouldn't that lead to a very skewed poll no matter how accurate their algorithms are?
I've looked all through their analysis (actually, "essay 1" and "essay 2" lead to the same thing - so maybe there is something missing), but I have not seen how they address this issue. Like I said, I'm sure their analysis is fine, but if they're basing it on polling trends, and those are skewed, then is it truly accurate.
I also read Anonymous' Fox News article, and I've seen the same info spread around the web: the skews in the polls are very unrealistic. Take the CBS/NYT/Quinnipiac poll for VA from the 19th of September. The sample was 35% D 24% R. That means that there are 46% more democrats than republicans in the state. This is just one of many examples, but there is no data out there whatsoever that shows VA's party affiliation is anywhere close to this. To make it even more flat, the media reports a 4 point lead for Obama, but, if the sample was even Ds and Rs, if D was oversampled by 11 percentage points, wouldn't it be safe to assume he should be up by about 11 in the polls? Voter breakdown in that poll actually shows that Is vote heavily for Romney. No media outlet reports this, however. No one says the full truth of "Obama is up by 4 in VA; but we sampled 45% more democrats than republicans. Moreover, independents favored Mitt over Obama by 53% to 42%." In fact, in most of these polls, you'll actually notice that independents are leaning towards Mitt. It's not being reported this way, though. With respect to party affiliation in the states, from a sheer common sense point of view, voter enthusiasm and registration are down for democrats. By how much, who knows? But the fact remains that the layout of the 2012 election is not going to be anywhere close to that of a very statistically dissimilar 2008 election.
To me, all the algorithms are meaningless if the basis of the poll, eg. the makeup of the sample is skewed. Samples should reflect the makeup of the population. If VA was 38% D and 36% R, in order to get an accurate poll, that's what the poll should sample. But regardless, I cannot get around this egregious over sampling, and I can't get around why the media is reporting it how it does when there is a whole other side to tell.
I hope to hear back from you. If not, have a great weekend!
That article still concludes that Obama is ahead with a decent lead, even accounting for the combined margin of error and potential oversampling that the article made the case for.
DeleteAh, seems I hit reply to Anon just as you finished your post! Good morning/Happy Friday to you as well!
DeleteI certainly see your point in some poll numbers simply not adding up, particularly that VA example. I definitely can’t see how there would be 46% more Democrats in the state, but at the same time the poll results don’t necessarily imply that; it implies that there are 46% more Democrats likely to vote than there are Republicans likely to vote. That statistic seems far plausible; particularly considering the fact that Mitt hasn’t succeeded very well in rallying up the GOP behind him as a candidate.
Of course though, the same argument could be made that Democrats likely aren’t anywhere near as rallied up behind Obama as they were in the past either. Although, this is of course difficult to quantify accurately.
I understand the concern about them using all state polling data and not excluding any potential bias or poorly done polls, but at the same time if the sample size is large enough then the worst polls should just end up as statistical anomalies/outliers. Bad polling certainly isn’t unique to this election, and knowing that the methods applied here still were able to very accurately predict the past two presidential elections.
I agree with you though that these factors aren’t making the news when they should be. Reporting on any poll like that as if it’s gold, should be accompanied with a breakdown of the statistics/criticisms, but of course that’s apparently asking for far too much from media.
- The man behind the anti-Muslim film "The Innocence of Muslims" was held without bond Thursday evening on charges that were not related to his film -
ReplyDeleteI'm confused here. Yes. The guy 'allegedly' violated his parole. Yes. The courts have a right/duty to revoke said parole. But, is he not still 'innocent until proven guilty'? And if he WAS revoked, shouldn't he simply be remanded to serve the remainder of his sentence... which was 21 months from his 2010 conviction?... and why were no reporters allowed in the courtroom for a mere parole violation hearing? Held w/o bail b/c he's a 'flight risk'? Where's he going to go? Certainly NOT back to Egypt... due to MSM's outing of his photo/identity/address - he's got a price on his head, here in America, as well.
EVERY report I've seen has stated up front... oh noes!... 'this is totally unrelated to the film'... and then there's this - slated to happen after Friday prayers in Dearborn, MI the day AFTER the guy is picked up, arrested and held w/o bond:
http://www.breitbart.com/Big-Peace/2012/09/27/Blasphemy-Laws-The-Next-Step-in-Religious-Tolerance?utm_source=twitterfeed&utm_medium=twitter
utm_campaign=Feed%3A+BreitbartFeed+%28Breitbart+Feed%29
I find it sort of funny (not the ha!ha! kind) when immigrants use THEIR Constitutional First Amendment right to peaceful assembly to protest OUR Constitutional First Amendment right of free speech - demanding it be censored with regard to THEIR beliefs while THEY continue to 'blaspheme' OUR beliefs, in total compliance with the First Amendment...
Amazing world we live in today!