Haselton needs an article about basic statistics. The 95% confidence interval on the difference between the two proportions is 6% +/- 17%, i.e. the range from about -11% to +23%. This (a) demonstrates that the sample is indeed underpowered to distinguish the sort of effect sizes that Haselton appears to be interested in, and (b) demonstrates that a +20% difference in proportion, contrary to Haselton's assertion, absolutely falls within the range of true values that can't be ruled out at a standard level of statistical confidence given the outcome of this experiment.

see http://www.kean.edu/~fosborne/bstat/06d2pop.html for the basic statistics.

Unfortunately I don't have the superior powers of the autistic mind to catalog and recollect all this detail about arcane topics, be they movie producers or minor bugs in cellphone software.

Recently I met a gentleman whose profession was gathering up shopping carts in the supermarket parking lot, and he treated me to an extended discourse about the relative merits of the producers of the different Muppet movies. I recognized the classic signs of an autism spectrum disorder ("For example, a person with AS may engage in a one-sided, long-winded speech about a favorite topic, while misunderstanding or not recognizing the listener's feelings or reactions, such as a wish to change the topic of talk or end the interaction."), and for some reason, my thoughts turned to our old friend, Bennett Haselton.

Um yeah, I think this is easy to explain

http://en.wikipedia.org/wiki/Regression_toward_the_mean

One thing that does it (nexus 4/chrome) is a "fling" quickly followed by a touch. Normally the second gesture "catches" the inertial scroll and stops it, but in the slashdot mobile site it tends to be registered as a touch on whatever was scrolling past the finger. I think the same thing sometimes happens for two flings in a row (when it seemed like the first didnt register, but it was just laggy) or a fast fling followed by a touch & hold or slow drag.

Sam Wang is a professor of neurobiology, not statistics. Also, the article does not refer to his predictions but to those of Nate Silver, who predicts only an 86% chance of Obama winning, notwithstanding the incorrect calculations using his data in the random anonymous script linked to the story.

The 538 website publishes the marginal probabilities of each state's outcome. The random anonymous script that is linked in this story just takes the product of these to compute the joint probability of Obama winning a particular set of states. This is of course a mistake. The probability that Obama wins Pennsylvania and Ohio is not the product of the probability that he wins each state separately, unless those two events are statistically independent. Of course, in reality and in the 538 model, they are not -- if Obama loses Pennsylvania he is also more likely to lose Ohio. I think this mainly accounts for the difference between the 538 prediction and the "prediction" of the random anonymous crap that the story links.