Yesterday, Drew Conway posted an analysis of the survival time to events on Chatroulette. If you're familiar with Chatroulette, you'll know what kind of events you can expect to occur when using it. (If you're not, here's a hint: don't try it now if you're at work.)
Sadly, it was all an April Fool's Day joke. But Drew takes the opportunity to teach an important lesson: it's really really easy to simulate some data (Drew provides all the necessary code to do so in R), create some convincing-looking charts, and publish an academic-sounding review of a "paper". If Drew's article hadn't been posted on April 1, I'd bet good money that a news site somewhere would have linked to it.
Drew's data was simulated, but Casey Neistat did actually collect real data (example: average time-to-pervert) from a Chatroulette session, and even did a controlled experiment (time-to-next for guy versus pretty girl). He made a movie about it, it's kinda cute (especially the data presentation) but also slightly NSFW. Enjoy.