Re: [FRIAM] Poll: Obama leads all GOP candidates in head-to-head contests
Very nice aggregation .. and interesting. Agreed bias does show through. I looks like a potentially very close general election. Especially if the former article is right: once a republican candidate is chosen, the gap closes. This sounds like a reasonable hypothesis. -- Owen On Wed, Mar 14, 2012 at 12:09 PM, glen g...@ropella.name wrote: Personally, I find polls confusing. I like this page: http://www.realclearpolitics.com/epolls/2012/president/president_obama_vs_republican_candidates.html that projects multiple polls down to a simpler space. Best I can tell, the error for the ABC/WaPo poll is is 4 points, the Pew poll is 3 points, and Rasmussen's is 3 points. Even ignoring how the questions are worded and asked and their domains (RV/LV, demographics, geography, etc.), we see a lot of wiggle from poll to poll. The most interesting thing to me are the correlations between who's responsible for the poll, their political bias, and the results of the poll. It would be interesting to see how much and in what direction each wiggles over time. E.g. does Rasmussen wiggle _more_ than Pew? Or do WaPo polls wiggle with a skewed distribution? It's obvious that statistics are lies. But perhaps the statistics of the statistics would show patterns in they lying that would allow one to spot lying trends. FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
Re: [FRIAM] Poll: Obama leads all GOP candidates in head-to-head contests
George Box gave us: All models are wrong, some are useful and he was speaking specifically of *statistical models*... I think the rest of us apply it a bit more widely, and I don't think that is inappropriate. A corollary to this might be, all statistics are wrong, some are useful? I'm not sure Statistics are Lies is precisely accurate. I think Statistics are incomplete and Statistics are skewed come closer but *even* more to the point, I think, is Statistics are used to lie. Politics is unlike science in that it *always* seeks to persuade rather than illuminate. To the extent that statistics (or a model) illuminates a process or phenomena it is useful. Journalism (in principle) falls more close to Science, attempting to *Illuminate* rather than *persuade*, but in fact, I don't know that much objective journalism is practiced these days... much journalism is thinly disguised politics or at least heavily informed by the ideology of publisher/editor/backers. I'm not attacking journalism as conceived, but rather more as practiced. I'm not sure where Polling falls. In principle, it is based in sound scientific/statistical principles, but in fact it is funded/driven either by *overt* political interests (parties, PACs, etc.) or *covert* ones (biased media). To the extent that news and politics are a coupled, self-referential system capable of feedback loops in near real time, it should not be surprising that some folks in the trade (either J or P actually) may choose to declare their opinions (or more to the point preferences or fantasies) as facts, hoping/trusting they will be consumed as such. Where more justification is required, it seems easy enough to deliberately skew polls (in their formation or in their analysis or both) in a way that is far from honest but maybe not evidently so upon casual observation. This leads me back to Americans Elect who have been discussed here off and on. My initial turnon was their superficial rhetoric... it sounded like they wanted to help me take back my vote from the conceptually gerrymandered process that currently seems contrived specifically to make sure I only have lesser of evils to vote for. When I signed up and filled out their questionaire, etc. I was appalled at the nature of the questions and the multiple choices allowed for answers. All were at best half-truths and many were very misleading IMO... much like a poll being taken during a heated campaign. I'm just now dipping back into Americans Elect... it seems more positive than I remember over (nearly?) a year ago when I first dipped in... 30+ questions into their match questions I am only skipping or unsuring about 1/10. An odd artifact is the way Wyoming seems to be an outlier in their answers to many questions... ND is another... maybe just low sampling? Once again... statistics about the statistics would hlep! - Steve Personally, I find polls confusing. I like this page: http://www.realclearpolitics.com/epolls/2012/president/president_obama_vs_republican_candidates.html that projects multiple polls down to a simpler space. Best I can tell, the error for the ABC/WaPo poll is is 4 points, the Pew poll is 3 points, and Rasmussen's is 3 points. Even ignoring how the questions are worded and asked and their domains (RV/LV, demographics, geography, etc.), we see a lot of wiggle from poll to poll. The most interesting thing to me are the correlations between who's responsible for the poll, their political bias, and the results of the poll. It would be interesting to see how much and in what direction each wiggles over time. E.g. does Rasmussen wiggle _more_ than Pew? Or do WaPo polls wiggle with a skewed distribution? It's obvious that statistics are lies. But perhaps the statistics of the statistics would show patterns in they lying that would allow one to spot lying trends. Owen Densmore wrote at 03/14/2012 10:32 AM: This: http://news.yahoo.com/blogs/ticket/poll-obama-leads-gop-candidates-head-head-contests-151032764.html is an interesting Pew poll on the whole election scene, Republican primary as well as an Obama second term. I'm a bit surprised about the Health Care tie, although is makes sense if one compares other alternatives that at least approach a single payer. -- Owen Quote: President Barack Obama is leading all of the Republican presidential candidates in head-to-head match-ups, according to a poll released Wednesday by the Pew Research Center for the People and the Press. http://www.people-press.org/2012/03/14/romney-leads-gop-contest-trails-in-matchup-with-obama/ A national survey taken March 7-11 showed Obama leading Romney by 12 percentage points (54-42) and even further ahead of Santorum with 57 percent of support to Santorum's 29 percent. Those numbers are likely to shift as Republicans rally around a single candidate in the coming months, but as a
Re: [FRIAM] Poll: Obama leads all GOP candidates in head-to-head contests
Steve Smith wrote at 03/14/2012 12:34 PM: I'm not sure Statistics are Lies is precisely accurate. I think Statistics are incomplete and Statistics are skewed come closer but *even* more to the point, I think, is Statistics are used to lie. I can't resist. ;-) The noun lie is interesting. It's not like, say, hammer ... or rock ... well, unless you're a fan of intelligent design, that is. A lie is a thing that one might find lying [ahem] around on the ground. But somehow we can know just by looking at the lie, the purpose to which the lie was put. A lie is more like a hammer than a rock, of course. Those of us with hands (or with the neural structures that allow us to imagine hands) can accurately infer the purpose to which a hammer was put. The set of observers capable of accurately inferring the purpose of a hammer is quite small, but still seems large enough to those of us in the set. It's not so easy with rock. Was it designed to filter water? ... or to execute people who break your laws? -- glen FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
Re: [FRIAM] Poll: Obama leads all GOP candidates in head-to-head contests
Since politics seems to be rearing its ugly head here, I would like to recommend a book: Red State, Blue State, Rich State, Poor State, by Andrew Gelman. He is a well-respected statistician and political scientist at Columbia University. To quote from his web page: Andrew has done research on a wide range of topics, including: .. why campaign polls are so variable when elections are so predictable. Joe On 3/14/12 12:52 PM, Owen Densmore wrote: Very nice aggregation .. and interesting. Agreed bias does show through. I looks like a potentially very close general election. Especially if the former article is right: once a republican candidate is chosen, the gap closes. This sounds like a reasonable hypothesis. -- Owen On Wed, Mar 14, 2012 at 12:09 PM, glen g...@ropella.name mailto:g...@ropella.name wrote: Personally, I find polls confusing. I like this page: http://www.realclearpolitics.com/epolls/2012/president/president_obama_vs_republican_candidates.html that projects multiple polls down to a simpler space. Best I can tell, the error for the ABC/WaPo poll is is 4 points, the Pew poll is 3 points, and Rasmussen's is 3 points. Even ignoring how the questions are worded and asked and their domains (RV/LV, demographics, geography, etc.), we see a lot of wiggle from poll to poll. The most interesting thing to me are the correlations between who's responsible for the poll, their political bias, and the results of the poll. It would be interesting to see how much and in what direction each wiggles over time. E.g. does Rasmussen wiggle _more_ than Pew? Or do WaPo polls wiggle with a skewed distribution? It's obvious that statistics are lies. But perhaps the statistics of the statistics would show patterns in they lying that would allow one to spot lying trends. FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org -- Sunlight is the best disinfectant. -- Supreme Court Justice Louis D. Brandeis, 1913. FRIAM Applied Complexity Group listserv Meets Fridays 9a-11:30 at cafe at St. John's College lectures, archives, unsubscribe, maps at http://www.friam.org
