tag:blogger.com,1999:blog-6894866515532737257.post8386677981747695141..comments2024-03-27T01:01:09.785-07:00Comments on Probably Overthinking It: Proofiness and electionsAllen Downeyhttp://www.blogger.com/profile/01633071333405221858noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-6894866515532737257.post-46134659329321860422016-11-29T18:25:50.126-08:002016-11-29T18:25:50.126-08:00^_^^_^Anonymoushttps://www.blogger.com/profile/11113984241425650034noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-50178042642302586312016-11-29T08:24:53.292-08:002016-11-29T08:24:53.292-08:00I think your 41% differs from my 30% because you a...I think your 41% differs from my 30% because you actually computed your answer, while I just made mine up as an example. I'm actually mildly surprised that I was as close as I was. Allen Downeyhttps://www.blogger.com/profile/01633071333405221858noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-3364746769493673162015-01-08T22:57:55.516-08:002015-01-08T22:57:55.516-08:00I assume that the distribution of Franken's vo...I assume that the distribution of Franken's votes follows Poisson distribution with lamda = 1,212,629, and Coleman's distribution is similar. Then I sample from the two Poisson distributions for 1000 times and compare the votes. The result is about 41% for Coleman to win. Is there something wrong with my assumptions? Why my result differs from yours? Thank you.Anonymoushttps://www.blogger.com/profile/11113984241425650034noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-35097972958458634852012-02-04T15:24:02.623-08:002012-02-04T15:24:02.623-08:00Yes, I am suggesting that the outcome should alway...Yes, I am suggesting that the outcome should always be determined by a biased coin toss, but to be clear, the bias would be based on posterior probabilities, not the fraction of ballots. With the very large sample sizes in most elections, the chance of a clearly disfavored candidate winning would be astronomically small.<br /><br />The bias would be effectively 0 (or 1) for any outcome other than a statistical tie. But the advantage of what I am proposing is a smooth transition from A wins to tie to B wins, without any thresholds that would be a point of contention.Allen Downeyhttps://www.blogger.com/profile/01633071333405221858noreply@blogger.comtag:blogger.com,1999:blog-6894866515532737257.post-90192613577989879442012-02-04T00:42:10.815-08:002012-02-04T00:42:10.815-08:00Do you suggest always flipping a biased coin? If y...Do you suggest always flipping a biased coin? If yes, there's the unsettling possibility of getting a clearly disfavored candidate. Otherwise the boundary between "decided" and "contested" elections is still just as arbitrary as with Seife's coin flip. I'm not sure there's any authoritative way to set a margin of error, except maybe to commit to one ahead of time based on past experience.Anonymousnoreply@blogger.com