I do not think that word means what you think it means

As reported in the Financial Times:

"We were seeing things that were 25-standard deviation moves, several days in a row,” said David Viniar, Goldman’s chief financial officer.

By Chebyshev's inequality, no matter what the probability distribution, the fraction of points which lie within k standard deviations of the mean must be at least (1 - 1/k²). For a normal (Gaussian) distribution, the probability of a 25 standard deviation outlier is much less, so much so as to be negligible. But even for the less stringent bound, the probability of being 25 or more standard deviations from the mean is at most 1/625, or 0.0016. To do that for three days in a row is at most (1/625)³ = 1/244140625 = 4.096 * 10^-9.

Put another way: There are about 250 trading days in a year (weekdays, excluding holidays). Even at the Chebyshev limit, it would take an expected million years or so to get three straight days of 25 standard deviation outliers. And Viniar's comment does not rule out a longer run of outliers.

Either these guys have botched the calculation, or they do not understand what standard deviation actually means.

(h/t Brad DeLong)

Triple 2007s

If I've timed this right, the timestamp will be (in European format) 20-07-2007 20:07.

We'll get one of these a year until 2012, then not until 2101.

No AC

The geniuses who "designed" our air conditioning "system" have struck again.

Sunday afternoon there was a series of thunderstorms in the area. Nothing severe hit Durham, although Rochester was hit with a microburst.

Nonetheless, Sunday's storms are being blamed for toasting a relay in the chiller plant. As a result, we have no air conditioning until further notice. They don't know when a full-fledged replacement transformer will come in, so they are promising that they will try to get something rigged up--but that will probably not happen before the middle of next week.

Air conditioning system failures have long been an annual (or more often) occurrence here, but this is absurd. Being honest-to-goodness rocket scientists (yes, we build space flight hardware here), we know better than to design systems with easily avoidable single point failure modes, especially in critical components. This is exactly what our B&G people are claiming we have here: one component failed, supposedly during a non-severe thunderstorm (not exactly a rarity in the summer), and we have to sweat it out for more than a week.

I want to hope that the university didn't pay a lot for this system--after all, air conditioning systems have been around for decades--but I'm sure they paid too much.

You're giving away what?

I got an invitation in the mail, allegedly from The Donald, to a seminar being given by his son Eric in Portland on 27 July. The invitation waives the "suggested tuition fee" of $149 and claims that I would "learn what others have paid over $20,000 to learn."

I smell a scam. People just don't give away $20K of information like that to strangers. Especially when they involve "government approved investments guaranteeing 9.6% to 32% return." I'm no expert on investments, but I didn't just fall off a turnip truck, either.

One other oddity. The return address on the envelope gives an address in DC (zip code 20036), but the postmark gives only the zip code 84199, which is in Salt Lake City.

Slot screws are evil

A follow-up to yesterday's post.

I don't know why anybody still uses slot screws. There are so many better options out there: options which don't invite the driver to slide out and cause serious damage to people and/or property.

My contractor agrees with me that slot screws are evil. He says that the only time he ever uses a slot driver is to take out old screws so that he can replace them with something better.

M. C. Escher, design engineer

I tried to hang a new curtain rod at my house Sunday and Monday. Because the run is longer than 48 inches, I need to add a couple of support brackets. The problem is that the brackets cannot be installed before the curtain rod is in place, and (at least with the tools I have) the brackets cannot be installed while the curtain rod is in place.

Why do so many designers try to emulate M. C. Escher in products people are actually supposed to use?

Stupid newspaper tricks

Foster's Daily Democrat
333 Central Ave.
Dover, NH 03820

Dear Foster's:

Thank you for coming up with your brilliant marketing technique of giving free one week trial home delivery subscriptions to unsuspecting homeowners who may or may not be in town that week. This makes our jobs much easier.

--The burglars of the Seacoast region

This week several houses on my street, including mine, were "lucky" recipients of the aforementioned trial subscription from the local fishwrap, Foster's Daily Democrat [sic].

Now Dover is not Miami, so I can understand why it would take them a minute (as opposed to five seconds) to realize why this is a bad idea. But they do publish police blotter columns, and some fraction of the crimes reported therein are burglaries. What better way to let burglars know who is out of town than to let newspapers pile up at a house whose owner didn't know they would be piling up?

Oh, and another piece of free advice to Foster's: Don't give free advertising to the competition. Their papers are delivered in plastic bags which advertize home subscriptions to the Wall Street Journal. It just so happens that the Wall Street Journal's parent company owns the other daily newspaper in the Seacoast region, the Portsmouth Herald.

British mailing addresses

Brad DeLong, trying to subscribe to British newspaper Financial Times, asks:

Did that form from http://ft.com really just ask me to type in my "house name"?

From which I infer that Brad doesn't publish in Elsevier journals. Their complete postal address: Elsevier Science Ltd., The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom.

The language of science

For the last sixty years or so, almost all cutting edge science has been published in English, and I am lucky that English is my native language. There has been some interesting work published in Russian, especially during the Cold War, but almost all of the interesting work gets translated to English within a year or two, so not reading Russian is not a big disadvantage.

However, English has not always been the dominant language of science. Before 1940 or so, the language of choice would have been German. In earlier eras, the preferred language of scientists might have been French, Latin, Arabic, or Greek.

I was reminded of this recently when my literature search on wavelets led me to the original reference on the Haar wavelet. Alfred Haar, then a professor of mathematics at Göttingen, published his paper "Zur Theorie der orthogonalen Funktionensysteme" in Mathematische Annalen (v. 69, p. 331) in 1910. This paper was, as can be guessed from the title, written in German. Our library turned out to have a copy of this volume, and so I borrowed it and tried to read it (I took three semesters of German as an undergraduate).

It really is not easy to read a technical article in a second language. Certain common words have meanings in technical usage which can differ substantially from their ordinary meanings (English and German are alike in this way). One example: stetig can be translated in ordinary usage as "constant" or "steady" (the latter word is its English cognate), but in mathematics it specifically means "continuous".

As I said, English has since displaced German as the preferred language of science. Mathematische Annalen is still published, but these days most articles are in English.

Nor is it inevitable that English will always remain the dominant language of science. The rise of science in India need not change this status quo since English is the one language that is widely spoken throughout India (it plays a similar role to that of medieval church Latin in Renaissance Europe). But China might change things--probably not in my lifetime, but maybe in the lifetime of scientists born this decade.