Andrew Gelman's Blog

I played around with this about a year or so ago. It's a neat little program, but I'm not sure how practical it really is for fitting models in most cases. It could be useful for some very light exploratory work, but one would have to follow up with some more rigorous analysis. That said, Andrew Gelman posted his thoughts on it awhile back. http://www.stat.columbia.edu/~cook/movabletype/archives/2009/12/equati… http://www.stat.columbia.edu/~cook/movabletype/archives/2009/12/equati…

Andrew Gelman has a great blot post on this: http://www.stat.columbia.edu/~cook/movabletype/archives/2010/12/the_tr… In a nutshell, the problem arises from trying to estimate small effects and then screening results on statistical significance.

This is a bit more technical, but still somewhat on topix. Andrew Gelman did a great "criticism" of Bayesian analysis on April Fools 2008: http://www.stat.columbia.edu/~cook/movabletype/archives/2008/04/proble… Followed later by his response to himself with some other statisticians: http://www.stat.columbia.edu/~cook/movabletype/archives/2008/07/respon…

False. The data in this "infographic" is really best represented as a scatter plot, using much less ink, as shown in [1] and similar other blog posts. The lines are unnecessary and misleading. It's good that Tufte has made concepts like data-to-ink density popular. But I urge you to actually evaluate that value when you look at a graph and refrain from using it to merely signal your knowledge of the right buzzwords. [1]http://www.stat.columbia.edu/~cook/movabletype/archives/2009/12/health…

You appear to have your facts wrong. From the study[1]: > Eligible cases were all patients with a glioma or meningioma of the brain diagnosed in the study regions during study periods of 2–4 years between 2000 and 2004. So participants in the study lasted between 2 and 4 years. Since elevated rates of cancer were only found in the group with the highest usage[2], this means people with >=1640 hours of talk time, which means between 68 and 34 hours per month. Not two hours per month. And this is easy to sanity check; here's the study again: > Corresponding values for [median lifetime cumulative call time] glioma controls were 100 h lifetime, 2.5 h/month and about 2000 calls. So the median user used about 2.5 hours per month, not the heavy user. Now, let's hear them out a bit more: > Overall, no increase in risk of glioma or meningioma was observed with use of mobile phones. There were suggestions of an increased risk of glioma at the highest exposure levels, but biases and error prevent a causal interpretation. The possible effects of long-term heavy use of mobile phones require further investigation. It's also worth noting that you're using the word double somewhat duplicitously. Yes, the study found double the glioma rate in very heavy users (as discussed, much more than two hours per month), but it also found that the rest of cellphone users are less likely to get cancer than average. If we take your causal analysis seriously, then we should also take this…

"The fallacy is that you're searching for a theory in a pile of data, rather than forming a theory and running an experiment to support or disprove it." But looking at data is an important step in forming new hypotheses -- http://en.wikipedia.org/wiki/Exploratory_data_analysis . You may just want to then gather more data to independently check on your ideas. (And to go back to the runs example, isn't the problem that the announcers weren't willing to seriously consider an alternate hypothesis, ie, they weren't doing enough simultaneous estimation?) "Instead of running multiple AdWords variants each against multiple landing page variants each feeding a different website funnel, run just one experiment at a time, one variable at a time." I think this is bad advice. If there's an interaction between your variables, this will lead you to totally miss it. Even without interactions, it can be a more efficient use of resources to estimate the effects of multiple variables at a time: http://en.wikipedia.org/wiki/Factorial_experiment Also, in this context, it seems worth looking at http://www.stat.columbia.edu/~cook/movabletype/archives/2008/03/why_i_…

It does make a difference for people: http://www.stat.columbia.edu/~cook/movabletype/archives/2007/05/names_… My hunch is that it also makes a difference for products, but I don't have any data on that...

Here you go: http://www.stat.columbia.edu/~cook/movabletype/archives/2010/12/reinve… One of the comments on the original paper (the pubmed link that teraflop cites) did point out that it was "just" the trapezoidal rule, FWIW, so the only people one may want to find fault with were the who stopped at his paper and decided to cite it, rather than dig further.

On errors in the "abortion lowers crime" study http://isteve.blogspot.com/2005/11/im-shocked-shocked-to-see-this.html overview of freakonomics 2 and how it shouldn't have been surprising that the errors were numerous and more obvious than in the original, due to the authors being less familiar with the topics in addition to selection bias http://www.stat.columbia.edu/~cook/movabletype/archives/2009/10/freako… Detailed discussion of blatant, blatant errors on their article refuting climate change. Makes you wonder if they knew their contrarian position was weak, and still put it in the book for controversy and the headlines it brings. http://climateprogress.org/2009/10/12/superfreakonomics-errors-levitt-…

from the article... the data is compelling ... Even their money fount, AdWords, is problematic. An illustrative anecdote: I ...did this and that... ... While I can’t quantify this, I’m confident that most engineers will agree: the article claims to argue from data, but presents no actual data: simply anecdotes. c.f., http://www.stat.columbia.edu/~cook/movabletype/archives/2011/01/for_th…

Not exactly the same variables, but similar enough: http://www.stat.columbia.edu/~cook/movabletype/archives/2007/11/religi…

Perhaps the sad fact is that being entertaining and being accurate are conflicting goals. Who wants to listen to a complex explanation for why the "fifteen minutes of videotaped conversation" method (http://en.wikipedia.org/wiki/John_Gottman )for predicting the likelihood of divorce is flawed due to overfitting issues (see http://www.stat.columbia.edu/~cook/movabletype/archives/2010/03/shooti… for details) Gladwell doesn't let complex facts get in the way of a good story.

What kind of exercise? Anecdotally, the kind that causes weight loss is pretty intense (i.e. sprinting, not jogging; Tabata is the most extreme form). See: http://www.stat.columbia.edu/~cook/movabletype/archives/2010/02/exerci…

Agreed. Here's the same data in a scatter plot (second chart down). Far, far easier to grasp. http://www.stat.columbia.edu/~cook/movabletype/archives/2009/12/health…

see http://www.stat.columbia.edu/~cook/movabletype/archives/2009/02/no_no_… "one possibility is that the more serious candidates (the ones we know are going to win anyway) are more attractive. Maybe you have some goofy-looking people who decide to run in districts where they don't have a chance, whereas the politicians who really have a shot at being in congress take the time to get their hair cut, etc."

Interesting enough book, but I must admit, I was surprised at the low quality of most of the popular videos and YouTubers he links from the "book". Are these people making any money out of their YouTube fame? I guess with a little dedication (produce content in regular intervals), you can still get quite far on YouTube. I feel reminded of "seeing faces on TV in the morning might help against depression": http://www.stat.columbia.edu/~cook/movabletype/archives/2007/02/watchi… Basically the guy found out that seeing talking faces in the morning lifts his mood. Perhaps that is all there is to the YouTube blogging phenomenon: people have a need to see talking faces, no matter what the faces actually say.

Perhaps this is why it generally takes a new generation of thinkers to push political thinking forward. I would postulate that we are more likely to reject the views of our parents than we are to reject our own. If that is true, it explains, for example, why gay marriage is a hot button issue for baby-boomers and older, but largely less of an issue for people under 30 (http://www.stat.columbia.edu/~cook/movabletype/archives/2009/06/future… I'll bet racism data looked pretty similar a few decades ago, but now expressing racist views publicly will get you in trouble even if you are extremely conservative (that Republican politician from the south a couple years ago who got extreme negative publicity for calling blacks "monkeys"). If it is necessary for one generation to die for political views to move forward, that doesn't bode well for progress if Aubrey de Grey and Ray Kurzweil prove prophetic. (http://www.youtube.com/watch?v=hgLRhxvRlKg) and (http://en.wikipedia.org/wiki/Predictions_made_by_Ray_Kurzweil)

Data on national comparisons of IQ are very poor. http://www.stat.columbia.edu/~cook/movabletype/archives/2010/04/jelte_… http://wicherts.socsci.uva.nl/wichertsPAIDrejoinder.pdf http://wicherts.socsci.uva.nl/wichertsRavenAfr2010rej.pdf http://wicherts.socsci.uva.nl/APA02.pdf This is a hard issue to analyze when most countries don't gather data on IQ that are comparable at all from country to country, and when there is considerable disagreement about what we mean by "religion." It is becoming a firmly established research finding that rationality is weakly correlated with IQ, http://www.project-syndicate.org/commentary/stanovich1/English http://www.oup.com/us/catalog/general/subject/Psychology/CognitivePsyc… with many high-IQ individuals exhibiting extremely poor rationality, so even if we knew differing IQ distributions for countries around the world reliably, which we do not, we would have no idea from those statistics which countries have a populace that tends to act rationally.