Your leaking thatched hut during the restoration of a pre-Enlightenment state.


Hello, my name is Judas Gutenberg and this is my blaag (pronounced as you would the vomit noise "hyroop-bleuach").


decay & ruin
Biosphere II
dead malls
Irving housing

got that wrong

appropriate tech
Arduino μcontrollers
Backwoods Home
Fractal antenna

fun social media stuff

(nobody does!)

Like my brownhouse:
   statistical eye
Tuesday, June 29 2010
This evening David was down in the city, so Penny came over with baby Milo, and the four of us had a dinner of vegan tapas (including tamales, spicy mushrooms, and some sort of potatoes-in-tomato-sauce concoction. As always for entertaining guests in this season, we ate out on the east deck. (We used to entertain on the south deck, but the east deck has proved considerably less plagued by mosquitos.) The evening was cool and breezy, with temperatures perfect for sitting outside. There was some leftover wine and Hite (a Korean beer) from Milo's first birthday party. Conversation mainly centerered around various ongoing social turmoils affecting first Penny and then Gretchen. At some point Ray showed up, having gotten out of work early. He'd brought a bottle of wine, but it was decided it was too much of a Sauvignon Blanc, so the rest of us babysat Milo while Penny drove home to get more wine and some movies she'd rented from a video rental place (perhaps from a place that also does black smithery and sends telegrams).
We tried to watch a movie called Crazy Heart, but Milo was making too much noise so eventually we abandoned the idea and Penny and Milo went home. (The original plan had been for them to spend the night; we'd even done a cleaning jihad in anticipation of this and other imminent sleepovers.)

I've been watching with interest the ongoing kerfuffle concerning Research 2000, a polling company contracted by DailyKos (the liberal megablog) to, among other things, provide large-font polling data on the Obama administration at the top of the DailyKos homepage. Recently, though, DailyKos has charged Research 2000 with fraud, saying (essentially) that the poll results they'd been providing were not the result of expensive call centers talking to actual random samples of the population, but were instead simply made up, presumably reflecting what the liberal writers and readers of DailyKos wanted to see. (You can laugh, but made up data is easy to assemble and can take you pretty far. In 1986 I won a Virginia state-wide science fair in the psychology category using made-up data about chicken responses to ærial alarms.)
But it turns out that it is difficult for human beings to create data that conforms to either real randomness or natural phenomena. For example, if one asks a person to sit down and write out the results of an imagined coin toss experiment, his data will be easy to distinguish from the data provided by someone doing an actual coin toss. This is because a person faking the data will not include as many runs of heads-heads-heads-heads-heads or tails-tails-tails-tails-tails as an actual coin toss produces. Similarly, then, if a polling outfit seeks to save money by simply making up its data, they'd better be good at it or it will be obvious, at least when inspected either by statistical methods or by someone with a keen eye for polling data.
Nate Silver is perhaps the most celebrated polling data genius in the business. At his website,, his analysis of polling results are nuanced and virtuosic, allowing him to tease accurate predictions from disparate data sources like some sort of oracle (or, rather, like an oracle would have you believe). Some weeks ago (or perhaps earlier) Silver had noted that the polling results from Research 2000 did not behave as other polls did. There were too many poll-to-poll changes (up or down a point) and a suspiciously-low number of unchanging polls (neither up nor down a point). In polls where the results hover near 50%, the number of days of unchanging polls should be the highest, followed on either side by the number of days where the poll changed either a point up or down. Further out, the number of points of change should taper off in both directions to trace a bell curve, that is, they should appear in a "normal distribution." But for, say, Obama's favorability ratings, Research 2000's distribution looked more like the cross section of Mount Mazama after the titanic explosion that created Crater Lake. To most mortals, the polls looked like polls, swinging convincingly up and down like polls do. But to Nate Silver's trained eye, the polls looked like a fraud, the result of data fabrication. It shows the difference between a virtuoso and a casual observer.
I was thinking about all this tonight while I was stoned on marijuana, the first time I'd used this illegal gateway drug since 4/20/2010. I realized something about marijuana tonight: when I'm stoned, I feel like my brain is an integral (as in the calculus operation) of what it is when I am not stoned. I start seeing patterns in my thought and the thoughts of others that I hadn't noticed before, perhaps because I was paying too much attention to their immediate content or context. To step back and take a meta-view, to see the underlying structures of thoughts and perceptions (even if it's all a drug-induced mirage), that seems useful somehow.
On a related note, my brother-in-law recently completed a government study (your tax dollars at work!) wherein he drew a statistical connection between marijuana use and the mention of marijuana in pop songs (also check out the take on a related study he did about music and sex by so-called "values voters"). The study failed, unfortunately, to tease out an answer to the most interesting question it could have tried to answer: does stoner music cause people to smoke pot, or does pot make people like stoner music? Personally I don't much like stoner music, but I do like to occasionally smoke pot by myself, think about things, and then masturbate.

For linking purposes this article's URL is:

previous | next