Along with Shelley, I am a graduate student in the Neuroscience Program at UM. The last three years my labmates and I have made a trilogy of satirical neuroscience posters poking mild fun at the mystical art of brain science. Shelley has kindly invited me to write on said trilogy. Also in any spare time remaining I punish myself with some rather difficult neural engineering experiments. (Tim Marzullo)
Episode 1: Spurious Correlations
You know the experience. To quote Allen Ginsburg, “everybody’s serious but me.” You walk around the massive meeting of the Society for Neuroscience, with 30,000 people, and everyone looks so much smarter, handsomer, and more “with it” than you. This is also called “imposter syndrome,” and it’s a common feeling among 1st and 2nd year graduate students. Such a large meeting can make your endeavors seem destructively minuscule.
Thus, I remember how relieving and refreshing it was at the meeting in New Orleans of 2003, as I was idly walking around the teaching neuroscience posters and a poster immediately caught my eye: Kai Schreiber’s fMRI (functional MECHANIC resonance imaging) gag study on how hitting a hammer to one’s forehead demonstrates the importance of the frontal cortex in decision-making. I thought the poster was hilarious as much for how out of place it as among serious education posters as for its content. I was immediately inspired to do a satirical poster of my own.
(Continued below the fold…)
A couple months later, as I was searching for ideas, a fellow neuroscience student made an off-hand comment at a departmental seminar about how he’d love to correlate fMRI brain imaging data to the activity of various stocks like Apple and Microsoft, so as to poke fun at how human imaging experiments often make heavy use of correlation. I jumped at it, saying I would help him out, and I pestered him over the next few months to give me some of his data so we could work on it and submit the work as a gag poster to the next SfN meeting.
However, he finally broke down saying he didn’t want to “shoot himself in the foot,” as it were. And he walked away. Me oh me, what do to do?
Screw it, I decided, I’ll just correlate the stock market to my own lab’s data! My initial reticence to using my own data was that I study rat motor cortex electrophysiology, and I didn’t think rats had much to do with the stock market.
But then I realized that the rats were precisely A MUCH BETTER REASON TO USE MY LAB’S DATA! Using human brain imaging data and correlating it to stock market data would just get lost in the noise at a the neuroscience conference. I mean, of course humans affect the stock market! But rat motor cortex neurons? Now that might be something remarkable.
I talked to my good friend and Matlab savvy labmate Greg on this idea. He didn’t even blink; he simply lightly grinned and said, “Yeah, we can do that, it will be fun.”
It turned out he had worked years before with a friend, Edward Rantze, from undergrad (Michigan State University, who knew?) on the idea of using some odd social sciences index factor to predict markets to invest in. His friend had already written software specifically to analyze day-to-day activity of the stock market and compare it to any input signal.
After running the activity of 8 days of rat motor cortex spiking data through Ed’s programs, we then had our results and proceeded to generate the poster in a hotel bar during the 2005 SfN conference in Washington D.C. It took all of two hours. Barflys were looking at us queerly, two dudes drinking multiple beers laughing out loud and fixated on a laptop computer.
The next day we found a Kinko’s nearby, printed our poster, and Greg put the poster up during the allotted time we were supposed to present the work. Oddly, according to Greg, a bunch of grad students from Berkeley were waiting for the poster! One student said, “We idly saw this in the abstracts section, and wondered if you were truly serious or not.” Greg, having done sales before he went to grad school, put on his game face (he was wearing a suit as well), and said we were a start-up company using the firing rates of neurons in the motor cortex of rats to predict the stock market as an alternative mutual fund for high risk/high pay off investors.
Some people became rather mad. One woman, who did DNA microanalysis, spoke in a rapid, elevated voice, “You’ve gotta correct for your number of samples. You can’t do this. Are you joking? This is so bullsh*t.” And then she stormed off. The poster spread around rather quickly; most people got the playful nature of the joke and enjoyed it. The best flattery came when I went down the escalators once, and I would heard two folks behind me, who I didn’t recognize, saying amongst each other, “there’s this really funny poster in the teaching neuroscience section correlating brain activity to stocks. Really Awesome. We should do one next year.”
I thought to myself, “Random people talking about my stuff?! Creating a buzz at the conference?! Hot Damn!” But….A Very Very bittersweet accomplishment. I wish my real neural engineering work, and not just my satirical work, generated such a response.
Over the next few months after the conference, we formally wrote up the study as a manuscript and submitted it to the Annals of Improbable Research. Marc Abrahams, who is the editor and also runs the Ig Nobel prizes, liked the manuscript; the work was published in the summer of 2006 and the cover art (above) made the back cover. It is available for viewing here.
Note: at this time my first manuscript in graduate school on the operant conditioning of the cingulate cortex had been mired in review difficulties for over a year. Way back in early 2003, a photographer from the magazine “MIT Technology Review” came to my lab and photographed our lab and rats for a feature of brain-machine interfaces. The first rat implanted for my study was featured in a photo in the article (picture below).
When the stock market work was published, the back cover of the magazine had some image-art of some juvenile rats that a friend and I had bred for fun. I used to make the joke, thank god it only lasted a few months, that my rats were more published than I was!
Final Technical Note: We debated calling the poster “How I learned to stop worrying and ignore the Bonferroni correction.” Since the New York Stock Exchange has ~5000 stocks, it follows statistical theory, given certain assumptions, that 5%, or 250, of those stocks will have a falsely significant relationship to any signal you choose. If we truly did this study correctly, we would divide our p-value (0.05) by the number of stocks (5000), at which point all of our results would become insignificant.
When I presented this poster, I would often ask people, and these are scientists mind you, if they could spot the error. Some could, many couldn’t.
But, to be fair, I probably wouldn’t have been able to spot the error myself if someone put me on the spot.