Study Hacks on Esther Duflo

Cal Newport, the guy behind my favorite grad student philosophy / self-help /how-to-be-awesome website, Study Hacks, just posted a pretty interesting article on Esther Duflo, MIT econ professor, vanguard of the randomization movement in development economics, recent Clark medalist, undergrad professor of Sol's, and all around awesome professor. Specifically, he uses her as an example of the do's and don't's of finding one's "Life's Mission" :

This is what complicates the mission to find a mission. On the one hand, to discover them (and recognize them), you need a non-conformist’s confidence and a dedication to exploration. Duflo, for example, was a notorious searcher. Among other acts of defiance, she took time off in the middle of her studies to go work on practical economic problems in Moscow (where she met Jeffery Sachs). When she took Banerjee’s class she was actively seeking an outlet for her intellectual energies.
On the other hand, this sense of exploration has to be backed with competence in the relevant field. And developing this competence has a decidedly unexciting, conformist feel to it — a process replete with hard focus and resistance to distraction.
What I find interesting about this (aside from the various compelling tidbits about Duflo's life) is how much I think it has in common with doing good interdisciplinary work. People love to think about doing work that spans fields, and at the same time a lot of the interdisciplinary work that's done is rightfully lambasted as being unrigorous. I think that's because to do it right one has to go through the drudgery of learning the ins and outs of not one but two fields. There's the flight of inspiration that comes from thinking up great cross-field paper topics ("Can we use monsoon strength as an instrument for trade costs?") but there's also the long stretches of building confidence in relevant areas ("Hm. Guess I better learn about the global climate circulation. And trade.")

Viewed in this light, one can think of Duflo's work as in a lot of ways combining the statistical techniques from epidemiology (or maybe even more rightly, pharmacology) with the concerns of development economics. Economists of my generation I think largely take for granted the progression towards randomized trials and emphasis on identification (though even that's changing fast, see for the example the Spring 2010 JEP), but the progress to incorporate those techniques involved taking the field and broadening its boundaries by mixing it with the techniques and concerns of another. It's nice to hear the back story about how one of the major players decided to head down that route and appreciate just how much skill, drive, and chutzpah it took to do so.

And yes, I just noticed that Study Hacks' byline is "Demystifying Sustainable Success." How apt.


Migrating birds on the front page of the NY Times

I'm very impressed and surprised by the coverage. And its a great story about innovations in scientific research. I didn't even realize that when I was a kid watching wildlife videos and learning about the Arctic tern that we couldn't actually observe them migrate. We could only see where they started and ended. But now with smarter tracking technology, we can observe their entire trajectory.

The whole story reminds me of other innovations in observational technology that slingshotted an entire field. For example, the invention of GFP, which can make portions of tissue glow, lead to enormous advances biology and related fields. (Martin Chalfie, one of the inventors who won the Nobel for it, is here at Columbia. I know because I saw him explain the idea to a gymnasium full of kids here with a [humorously] malfunctioning flash-light). I think that a lot of times, when we learn science in [grad]school, there is so much focus on theory, mechanisms and methods that we sometimes forget that the starting point of all science is observation.

PS. If you're like me and did a double take at the article's nonchalant statement about the groundbreaking technology
Geolocators ... just record changing light levels. If scientists can recapture birds carrying geolocators, they can retrieve the data from the devices and use sophisticated computer programs to figure out the location of the birds based on the rising and setting of the sun.
You should check out the website of the geolocator manufacturer Lotek, where they post scientific papers on the method. Here's an abstract from "An advance in geolocation by light" by P. A. Ekstrom (2004):
A new analysis of twilight predicts that for observations made in narrow-band blue light, the shape of the light curve (irradiance vs. sun elevation angle) between +3 and -5.DEG. (87 to 95.DEG. zenith angle) has a particular rigid shape not significantly affected by cloudiness, horizon details, atmospheric refraction or atmospheric dust loading. This shape is distinctive, can be located reliably in measured data, and provides a firm theoretical basis for animal geolocation by template-fitting to irradiance data. The resulting approach matches a theoretical model of the irradiance vs. time-of-day to the relevant portion of a given day's data, adjusting parameters for latitude, longitude, and cloudiness. In favorable cases, there is only one parameter choice that will fit well, and that choice becomes the position estimate. The entire process can proceed automatically in a tag. Theoretical estimates predict good accuracy over most of the year and most of the earth, with difficulties just on the winter side of equinox and near the equator. Polar regions are favorable whenever the sun crosses -5.DEG. to +3.DEG. elevation, and the method can yield useful results whenever the sun makes a significant excursion into that elevation range. Early results based on data taken on land at 48.DEG.N latitude confirm the predictions vs. season, and show promising performance when compared with earlier threshold-based methods.


The achievements and under-achievements of our species

A friend sent me this spectacular time lapse video of the space shuttle preparation in an email. if you haven't seen it, you can watch it right here.


NCDC: Warmest April Global Temperature on Record

GISS was already saying that 2010 would likely be the warmest year on record a while ago due partly to the current El Niño (see, for example, section 6 in the Current GISS Global Surface Temperature Analysis from January) and observation data are now bearing this out. NOAA's NCDC just released their April State of the Climate report, and not only was April 2010 the warmest April on record, but 2010's Jan-Apr four month span was the warmest on record as well. A gridded temperature anomaly map is at right.

There's a couple of things about this that are interesting. The first, which relates to some of my own work looking at how people internalize new info about climate change, is that record events seem to be one of the most straightforward ways of conveying to people that the climate is changing. "Climate" is, after all, the general distribution of various properties of the ocean and atmosphere over time, not any specific value at any specific time. Difficulty with the notion that climate change is a shift in this distribution is what often seems to underlie opposition to the theory of anthropogenic climate change; this is what gives rise to, say, your uncle's comment that a cold winter day proves that Al Gore is a liar.

Records, however, seem to be one of the ways of talking about distributions that people intuitively understand. Everyone understands that Usain Bolt is by objective measures a very fast runner cause he holds the world record for the 100m and 200m dashes. Similarly, people intuitively understand that if this is the "hottest year on record" then the climate is in a hot state. Moreover, if we get several hottest years on record in a row, then the climate is probably getting hotter.

So what? Well, given the above it's probably not as much of a surprise that studies like Yale's F&ES Project on Climate Change have been finding a downward trend in beliefs about climate change. After a slew of record hot years the planet has has a host of second-and-third hottest years since 2005, not exactly the sort of news that makes headlines, even if by historical standards the globe is still hot. Should 2010 shape up, as Hansen has repeatedly predicted, to be the hottest year on record, we'll probably see just as many news stories talking about how "global warming is back" as we have recent stories about the "cooling trend" of the past few years.

This, in turn, makes for an interesting counterfactual political economy question: what if Copenhagen had been scheduled for this December rather than last? Not to put too-high a hope on the established mechanisms, but maybe our outcome would have been a lot more positive if delegates had arrived knowing that they had just lived through the hottest calendar year on record...


Time Lapse of Eyjafjallajokull Erupting

Want to see what hundreds of tons of aerosols being ejected into the atmosphere looks like in time lapse video? Of course you do. First commenter to suggest a way to use this as an instrument for something worthwhile gets all the fermented shark they can eat.

Hi, I'm Jesse, and I'm now blogging with Sol.

Iceland, Eyjafjallajökull - May 1st and 2nd, 2010 from Sean Stiegemeier on Vimeo.

Data transfer from Matlab to Stata and reverse

Zip file with Matlab code to export or import datasets to Stata. For use in conjunction with "insheet/outsheet". Importing data from Stata is restricted to numeric variables (but not the reverse).