Some weeks ago a graph went viral in the internet showing a correlation between chocolate and nobel prize winners per capita comparing countries. Among others the BBC commented on this, based on a study published in the the New England Journal of Medicine (Messerli 2012). Now, I am a big consumer of chocolate myself, so no harm done there. And it is clear that the whole exercise was meant to be fun. But apart from the obvious problem that a micro theory was supposedly be tested by macro data (see, for instance, here), it is also funny how the public seems to get hypnotized by statistics like little Mowgli by sneaky snake Kaa in the Jungle Book. Anyways, when I checked the relationship with the same source for the number of nobel prize winners, but different data for chocolate consumption, I don’t get anywhere near the perfect relationship of Messerli. (By the way, I am not claiming that my data is better, to the contrary, but it illustrates nicely the problem of lacking robustness in statistics.)
Lesson learnt: there are very few really robust relationships in social science, and the one between chocolate and nobel prize winners is, unfortunately, not among them…
Polidigwerkschdaeddle 
Well… sometimes the explanation why different datasets lead to different results are just too obvious: Looking at the graph, you would easily see that those nation who win noble prices, but do not eat chocolate, simply do not deserve it.