Friday, January 22, 2010

Tales from the arXiv: Apparently, Congress is a Sandpile

It's this type of research that gives complex systems a bad name. The authors "model" Congress using the sandpile model from self-organized criticality, and they even have a few lines mentioning Al Gore's praising of the sandpile model as evidence in their favor. I call bullshit. This is yet another one of those articles that makes me react with a resounding 'So what?'. Anyway, here are the details of the article:


Title: Stochastic modeling of Congress

Authors: M.V. Simkin, V.P. Roychowdhury

Abstract: We analyze the dynamics of growth of the number of congressmen supporting the resolution HR1207 to audit the Federal Reserve. The plot of the total number of co-sponsors as a function of time is of "Devil's staircase" type. The distribution of the numbers of new co-sponsors joining during a particular day (step height) follows a power law. The distribution of the length of intervals between additions of new co-sponsors (step length) also follows a power law. We use a modification of Bak-Tang-Wiesenfeld sandpile model to simulate the dynamics of Congress and obtain a good agreement with the data.

2 comments:

Jon said...

"How can it be bad analysis? It presents the data in such a manner that it follows a power law..."

I have enjoyed watching John Doyle demolish that.

Mason said...

They lost me when I noticed them quoting Al Gore's comments in support of their research. I figured that that was more interesting to point out than the usual power law bit (though admittedly they can sometimes be insightful). I certainly could have linked to one of the iterations of my power law rant, however. (Just remember that if you take enough logarithms that even Angelina Jolie can be turned into a straight line.)