Month: January 2010

Influential Statisticians

Seth Roberts9 Comments

This article (“Ten statisticians and their impacts for psychologists”) impressed me. It’s a lot more accessible and basic than the usual academic article. However, my list — of the statisticians who’ve had the biggest effect on how I analyze data — is much different than his. From more to less influential:

1. John Tukey. From Exploratory Data Analysis I learned to plot my data and to transform it. A Berkeley statistics professor once told me this book wasn’t important!

2. John Chambers. Main person behind S. I use R (open-source S) all the time.

3. Ross Ihaka and Robert Gentleman. Originators of R. R is much better than S: Fewer bugs, more commands, better price.

4. William Cleveland. Inventor of loess (local regression). I use loess all the time to summarize scatterplots.

5. Ronald Fisher. I do ANOVAs.

6. William Gosset. I do t tests.

My data analysis is 90% graphs, 10% numerical summaries (e.g., means) and statistical tests (e.g., ANOVA). Whereas most statistics texts are about 1% graphs, 99% numerical summaries and statistical tests.

Reflections on a Few Years of Blogging

Seth Roberts9 Comments

Andrew Gelman’s blog has lasted longer than this blog (and was responsible for this blog.) Recently Andrew looked back. It seemed like a good idea so I will follow his lead.

The two big surprises have been how easy it is and how helpful it is. In the beginning it wasn’t easy to find interesting things to say. Somehow it got easier and easier. Partly because I had more ideas — about omega-3, the umami hypothesis and fermented foods, the effect of animal fat on sleep. Partly because readers sent me interesting stuff. Partly because I started teaching at Tsinghua and moved to Beijing part of the year. Partly because the Shangri-La Diet produced results that I wanted to brag about. And — a very big part of it — because there are enough comments here and elsewhere to make me think people are reading it. I think everyone has an innate desire to be listened to. As our concerns and knowledge become more and more specialized, it becomes harder and harder to find an audience. When Spy magazine was around I read every issue three times. I was dying to talk about it with other fans. I couldn’t. I couldn’t find them.

Some of the stuff people have sent me has been incredibly helpful. Most of the examples involve trying my ideas. Taking omega-3 (via flaxseed oil or fish oil). Tyler Cowen’s experience, for example. Tim Lundeen’s results. The effect on sports injuries. Or eating more fermented food. Tucker Max’s experience. Not only does it make the whole subject much easier to talk about, it convinces me I’m on the right track. Some of the examples involve telling me about other more conventional data related to my ideas. For example, I’m very glad to know about hormesis, which supports my ideas about fermented food. Knowing about radiation hormesis makes me stop worrying about the small dose of radiation I get from my cell phone. The recent comment about two morning faces being better than one might turn out to be really helpful and important.

I haven’t read She Stoops to Conquer, an 18th century play, but the title is brilliant. My self-experimenation, I now think, had a dose of that because I was willing to do something as humble as study myself whereas most scientists wouldn’t stoop to that. Too low-status. Blogging has a lot of that. How many Berkeley professors blog? Uh, Brad DeLong? And someone else, rarely. Blogging is beneath them. Whereas half of Tsinghua students have blogs. They aren’t worried about appearing undignified. The phrase keeping up with the Joneses means your car has to be at least as expensive as your neighbor’s car, and so on. A kind of arms race. Such an arms race goes on in science: What you must do to appear high status takes up more and more of your resources, leaving less and less to actually make progress. So less and less progress is made. Self-experimentation breaks out of that vicious cycle. Blogging is the same thing more generally. Supposedly professors, especially at a place like Berkeley, have interesting things to say. But the demands of status, as Veblen described in the last chapter of The Theory of the Leisure Class, make it harder and harder for them to say them. Blogging breaks out of that vicious cycle.

When I taught introductory psychology I found I could often weave whatever I’d been thinking about into my next lecture. It’s good to start a lecture by saying “Something interesting happened to me a few days ago . . . ” Now I can just blog about it.

Psychophysics of Flavor Complexity

Seth Roberts15 Comments

If I need evidence that we like complex flavors, I will quote this passage from The New Yorker:

“This sauce is really good,” she said. “It’s so Jean-Georges. He does this French-and-Asian thing.” She warned me that she would need a few seconds to figure out its precise ingredients. (She refused to divulge them, on the ground that Vongerichten would consider the recipe “a trade secret.” I later learned from one of the waiters that the ingredients include powdered English mustard and soy sauce.) “It’s so complex,” she said. “It makes me smile.”

The soy sauce is fermented. As any regular reader of this blog knows, I believe we evolved to like complex flavors so that we would eat more bacteria-rich food. So we have something in our brain that measures complexity of smell/flavor and translates that into pleasure: the more complexity, the more pleasure.

My experience of cooking is that it isn’t easy to produce a lot of complexity using spices and stuff like garlic and ginger. It’s possible but not easy. Ordinary recipes, such as in Saveur, aim for a low level, with 5-8 spices. Chinese Five Spice has 5 spices; spice mixtures might have 8; curry powders might have 10. At Whole Foods, the ready-to-eat soups have twenty-odd ingredients. Apparently their soup designers don’t find it easy, either.

Then I discovered that miso by itself produced sufficient complexity. Miso soup doesn’t feel “under-complex”. Finally I understood why wine is such a powerful flavoring agent; wine, like miso, is fermented. It makes sense that foods that our complexity detector evolved to make us eat do a better job of setting off that detector than other foods.

Now consider how that detector works. Suppose you have two sources of sodium — two different salts, for example. You get the same saltiness from 2 g of Salt A as you do from 1 g of Salt A and 1 g of Salt B. I think complexity is quite different. I suspect that 2 g of Source A (e.g., miso) will produce a lot less complexity than 1 g of Source A and 1 g of Source B (e.g., wine).

I tried adding two fermented flavoring agents (miso and tsukudani) to soup. It worked! The result tasted clearly better than miso alone. Now I do this routinely. It’s very easy. The results have a level of deliciousness I can’t remember encountering before. Everything else I can eat (such as restaurant food) now seems less delicious. I think that three sources works better than two; whether four is noticeably better than three I don’t know.

The basic idea is there are strong sources of complexity (fermented foods) and weak ones (all other flavoring agents). One strong source = 10-20 weak sources. You get the best results by using several strong sources of complexity, perhaps three or more. Once you know this you no longer: 1. Obsess over recipe details (as in the New Yorker quote) because all complexity is alike and easily produced, just as no one worries about the source of saltiness. 2. Think traditional, time-honored recipes are better than what you can make yourself (e.g., Saveur). As far as I can tell food professionals (with one big exception) don’t understand this. I really enjoyed Top Chef Masters (a competition between 12 of the best chefs in America) but there was an almost total absence of fermented foods. Perhaps one chef used soy sauce. The winner, Rick Bayless, made a mole sauce. Mole sauces, which combine 20-odd weak sources of complexity, take hours. I think they produce less complexity than three fermented sources put together, which takes about a minute.

The Accidental Influential

Seth Roberts2 Comments

Duncan Watts, a Yahoo! researcher who studies networks, has some interesting things to say:

“If society is ready to embrace a trend, almost anyone can start one–and if it isn’t, then almost no one can,” Watts concludes. To succeed with a new product, it’s less a matter of finding the perfect hipster to infect and more a matter of gauging the public’s mood. Sure, there’ll always be a first mover in a trend. But since she generally stumbles into that role by chance, she is, in Watts’s terminology, an “accidental Influential.”

Epidemics and many other contagion phenomena have a power-law distribution (large frequency of small number infected, small frequency of large number infected). When my colleagues and I studied the distribution of rat bar-press durations, we found a power-law-like function where the “size” wasn’t number but duration. Most bar-presses were quite short; a few were quite long. We also found that expectation of reward had a big effect on the slope of the power-law function. I think Watts is saying that more attention should be paid to what determines the slopes of these power-law functions.

A recent article by Watts. Thanks to Hal Pashler.

The Door-in-the-Face Effect

Seth Roberts2 Comments

One of my Tsinghua students, a freshman, has been getting up early Saturday mornings to go to nearby Beijing University to attend a 4-hour intro psych class for graduate students. “What does the teacher talk about?” I asked. He showed me his notes. “The Door-in-the-Face Effect” was the heading of a little graph he’d drawn. “What’s that?” I asked. “If you get someone to help you in a little way, they’re more likely to help you in a big way later,” he said. I knew that result. It’s called the foot-in-the-door effect. “Your teacher made a mistake,” I said.

I was wrong. There is a door-in-the-face effect very similar to the foot-in-the-door effect. The door-in-the-face effect is after you make a big request that is turned down, you are more likely to get agreement to a small request.

Academic Horror Story (Harvard University)

Seth Roberts16 Comments

Are the heads of large companies worse than the rest of us? Aaron Swartz said as much when, in a discussion of Ken Auletta’s Googled, he called them “sociopaths”. Nicholson Baker seemed to have had similar thoughts when he said about the same book that “what Auletta mainly does is talk shop with C.E.O.’s, and that is the great strength of the book.”

Lawrence Summers, now in the Obama administration, was head of Harvard University, one of the world’s most powerful companies, from 2001 to 2006. Everyone knows about Summers’ repeated tendency to do the incredibly-inappropriate thing. A generous interpretation of those incidents is that Summers had lived a sheltered life. I believe they were signs of something much worse — signs of pathology — based on what he did to one of Harvard’s best employees:

Back in 2002, a new employee of Harvard University’s endowment manager named Iris Mack wrote a letter to the school’s president, Lawrence Summers, that would ultimately get her fired.

In the letter, dated May 12 of that year, Mack told Summers that she was “deeply troubled and surprised” by things she had seen in her new job as a quantitative analyst at Harvard Management Co.

She would go on to say, in later e-mails and conversations, that she felt the endowment was taking on too much risk in derivatives investments, and that she suspected some of her colleagues were engaging in insider trading, according to a separate letter written by her lawyer that summarized the correspondence.

On July 2 Mack was fired. But six years later, the kinds of investments she allegedly warned about did blow up on Harvard. The endowment plunged 22 percent last summer, in part due to the collapse of the credit markets. . . .

Mack, who holds a doctorate in mathematics from Harvard, had been with Harvard Management for just four months when she approached Summers. She asked him to keep her communications confidential, or risk making her life “a living hell.”

But on July 1, Mack was called into a meeting by her boss, Jack Meyer, then the head of Harvard Management.

The next day Meyer fired her, according to the letter from her attorney, Jonathan Margolis, a copy of which was obtained by the Globe. Meyer told Mack that she was fired for making “baseless allegations against HMC to individuals outside of HMC,” according to the Margolis letter.

Mack writes to Summers, alerting him to behavior by her co-workers that she believed could (and eventually did) have a very bad effect on Harvard. Fearing loss of her job, she asks him to keep her warning confidential. Summers fails to honor her request. What distinguishes this particular horrible behavior from more conventional examples of horrible behavior by incredibly powerful people is that Summers’ action did him no good. He didn’t backstab Mack to get to the top. He was at the top. He didn’t exploit Mack. He didn’t cheat Mack. This is coming across a courageous decent far-seeing person, much less powerful than you, who is trying to help you and all the people in your care . . . and giving that person a good hard kick. For no reason. There is something very wrong with Lawrence Summers.

Frontline’s recent show The Warning tells how Brooksley Born, when she was head of the Commodity Futures Trading Commission (1996-1999), did her best to protect the rest of us from exactly what Mack warned about. Summers told her, according to a third party, “you’re going to cause the worst financial crisis since the end of World War II. I have 13 bankers in my office that have informed me of this. Stop. Right away.”