Tag Archives: linear regression

Ocular regression

A phrase I learned from Aaron Clauset’s great colloquium on the non-ubiquity of scale-free networks.  “Ocular regression” is the practice of squinting at the data until it looks linear.


On the Iranian election returns, in Slate

In today’s Slate I write about the claim that the official Iranian election returns are too linear to be true.

The graph (via Tehran Bureau) looks pretty amaing; but in fact, as I explain, it’s pretty much what you’d expect real election data to look like.

One point there wasn’t room for in the piece; if you look carefully at the chart above, you’ll see that the folks at Tehran Bureau got the election returns to fit the line y = 0.5238x – 742642 very well.  But in some sense that’s irrelevant, unless there’s some reasonable expectation that  clerical powers-that-be would want faked election numbers to follow a funny line with a negative y-intercept.  When R.A. Fisher went after Gregor Mendel, it wasn’t just because Mendel’s results looked suspiciously regular; it was because they looked suspiciously close to Mendel’s theoretical predictions.  If Mendel shaded the data, consciously or not, that’s the direction it would go.

I mean, I can fit a really nice quadratic in x to the Iranian election data — or, for that matter, U.S. election data — but absent any reason to posit a vast parabola-loving conspiracy, it’s just not that suspicious.

Update: (June 18)  Lots more material around the web about Iranian election stats.  A preprint on the arXiv claims the official numbers violate Benford’s law, but Andrew Gelman says no. On the other hand, via Mark Blumenthal at Pollster, I learn that Walter Mebane at Michigan finds some suspicious-looking irregularities in the town-level data.

Tagged , , ,
%d bloggers like this: