So, this post is lots of fun, and I want to encourage folks to do more blogging where they take a bunch of data, graph it up, and test hypotheses with it. Scien

So, this post is lots of fun, and I want to encourage folks to do more blogging where they take a bunch of data, graph it up, and test hypotheses with it. Science is great! On the other hand, though I rarely cite Zed Shaw, I have a special fondness for the theme of his classic article "Programmers Need to Learn Statistics or I Will Kill Them All": http://zedshaw.com/essays/programmer_stats.html http://zedshaw.com/essays/programmer_stats.html ... and I used to review journal articles, so I can't resist nitpicking by pointing out that some of these conclusions are a lot more dubious than others. For example, this sentence: For women, a smile isn’t strictly better: she actually gets the most messages by flirting directly into the camera... Appears to be based on the difference between 9.1 new monthly contacts and 9.3 new monthly contacts in each category. Is that statistically significant? I rather doubt it. The correct conclusion here is almost certainly "making a flirty face is not significantly more effective than smiling, and you have to be sure to do it into the camera or the result will be a disaster, so smiling might be the more reliable tactic." Of course, if you really want to start picking nits you can bring up the need for some kind of multifactor analysis to correct for the fact that certain kinds of pictures are probably correlated with certain kinds of personalities. But that would go beyond what I understand. And it's not actually necessary to try for a Ph.D. every time you throw up a graph or two. Good fun! Publish more raw data and we can turn this into an great statistics problem set.