Social Science Statistics Blog

30 August 2010

Rigor and modeling in economics

In a postscript, Andrew Gelman laments a general trend he notices in economics:

My only real problem with it is that when discussing data analysis, [the authors] pretty much ignore the statistical literature and just look at econometrics. In the long run, that's fine--any relevant developments in statistics should eventually make their way over to the econometrics literature. But for now I think it's a drawback in that it encourages a focus on theory and testing rather than modeling and scientific understanding.

Gelman has an idea about why this might be the case:

The problem, I think, is that they (like many economists) think of statistical methods not as a tool for learning but as a tool for rigor. So they gravitate toward math-heavy methods based on testing, asymptotics, and abstract theories, rather than toward complex modeling. The result is a disconnect between statistical methods and applied goals.

There is likely a balance here that Gelman misses between theoretical modeling and statistical modeling. Economists are in the business of testing complex theoretical models. A complex statistical model may draw attention away from that narrow goal.

Not that I necessarily endorse that viewpoint. It simply feels slightly unfair to economists to say that their spartan statistical modeling is a product of their obsession with technical rigor.

Posted by Matt Blackwell at 9:43 AM | Comments (0)

27 August 2010

Jackman on the Australian Elections

If you enjoy Australian politics, betting markets, and sharp statistical analysis, take a look at Simon Jackman's blog. He has been killing it lately.

Posted by Matt Blackwell at 10:39 AM | Comments (1)

The Seven Deadly Sins of Contemporary Quantitative Analysis

You may think you have good reasons to not stop what you are doing and read Phil Schrodt's essay on the "Seven Deadly Sins of Contemporary Quantitative Political Analysis". But you do not. Not only does the piece make several astute points about the current practice of quantitative social science (in a highly enjoyable way, I might add), but it also reviews developments in the philosophy of science that have led us here. The entirety is excellent, so picking out an excerpt is difficult, but here is his summary of our current philosophical messiness:

I will start by stepping back and taking a [decidedly] bird's eye (Thor's eye?) view of where we are in terms of the philosophy of science that lies beneath the quantitative analysis agenda, in the hope that knowing how we got here will help to point the way forward. In a nutshell, I think we are currently stuck with an incomplete philosophical framework inherited (along with a lot of useful ideas) from the logical positivists, combined with a philosophically incoherent approach adopted from frequentism. The way out is a combination of renewing interest in the logical positivist agenda, with suitable updating for 21st century understandings of stochastic approaches, and with a focus on the social sciences more generally. Much of this work has been done last decade or so in the qualitative and multi-methods community but not, curiously, in the quantitative community. The quantitative community does, however, provide quite unambiguously the Bayesian alternative to frequentism, which in turn solves most of the current contradictions in frequentism which we somehow--believing six impossible things before breakfast--persuade our students are not contradictions. But we need to systematically incorporate the Bayesian approach into our pedagogy. In short, we may be in a swamp at the moment, but the way out is relatively clear.

His section on "prediction versus explanation" is also quite insightful and deserves more attention. The upshot:

...the point is that distinguishing scientific explanation from mythical (or other non-scientific, such as Freudian) explanation is one of the central themes for the logical positivists. In the absence of prediction, it cannot be done.

Warning: if you truly love significance tests, you might feel a little heartbroken when reading this essay.

Posted by Matt Blackwell at 9:34 AM | Comments (2)

28 May 2010

Jaywalking allowed?

David Brooks has a column in today's New York Times about our difficulties assessing risk, in light of the oil leak in the Gulf of Mexico. One tendency he highlights is our excessive faith in devices aimed at minimizing risk. Although his general point is probably correct, I think he makes a statistical error in the example he uses to illustrate this tendency. Brooks seems to confuse numbers with rates.

He writes: "More pedestrians die in crosswalks than when jay-walking. That's because they have a false sense of security in crosswalks and are less likely to look both ways." I think that more pedestrians die in crosswalks than when jaywalking because more pedestrians cross the street in crosswalks than anywhere else. If the rate of dying is higher for pedestrians in crosswalks than when jaywalking, we might attribute the difference to our overconfidence in safety devices. However, the rate of dying need not be higher in crosswalks than elsewhere for more pedestrians to die in crosswalks. As long as many more people cross in crosswalks, it is likely that more will die in crosswalks. Brooks' point about risk assessment would have been stronger if he had considered not only the numerator (the number of deaths) but also the denominator (the number of people at risk of death) of the rate of pedestrian deaths that he's interested in.

Posted by Deirdre Bloome at 3:03 PM

19 May 2010

Are you friends with your friend's friend?

Measuring the extent to which our peers influence our behavior is hard for many reasons: one of the most basic is the difficulty of measuring who is a peer.

View image

Continue reading "Are you friends with your friend's friend?"

Posted by Deirdre Bloome at 11:37 AM

7 May 2010

Mathematical Sociology

James Montgomery, University of Wisconsin sociologist and economist, has a draft of a new book on mathematical sociology available for download. Unlike James Coleman's classic 1964 text, Montgomery's comes complete with Matlab code --- a sign of 45 years of progress :)

Posted by Deirdre Bloome at 5:26 PM

16 April 2010

Hainmueller in the New York Times

Jens Hainmueller, Assistant Professor at MIT and former writer for this very blog, has had some of his research written up in the New York Times today:

"Americans, whether they are rich or poor, are much more in favor of high-skilled immigrants," said Jens Hainmueller, a political scientist at M.I.T. and co-author of a survey of attitudes toward immigration with Michael J. Hiscox, professor of government at Harvard. The survey of 1,600 adults, which examined the reasons for anti-immigration sentiment in the United States, was published in February in American Political Science Review, a peer-reviewed journal.

There is an ungated version of the original paper.

Posted by Matt Blackwell at 8:36 AM

15 April 2010

The inevitable R backlash

There is a blog post floating around by Dr. AnnaMaria De Mars, where she speculates on what the "next big thing" is going to be. Apparently, it is data visualization and analyzing unstructured data, but not R:

Contrary to what some people seem to think, R is definitely not the next big thing, either. I am always surprised when people ask me why I think that, because to my mind it is obvious...I know that R is free and I am actually a Unix fan and think Open Source software is a great idea. However, for me personally and for most users, both individual and organizational, the much greater cost of software is the time it takes to install it, maintain it, learn it and document it. On that, R is an epic fail. It does NOT fit with the way the vast majority of people in the world use computers. The vast majority of people are NOT programmers. They are used to looking at things and clicking on things.

(I am not sure how a "non-programmer" is going to be able to analyze unstructured data or create wonderful visualizations, but that is beside the point.)

The ease-of-use argument or the "not everyone is a programmer" argument is one to which I am sympathetic. It has become quite heated in the debate over the Apple iPad in the last few months. Where the iPad succeeds is to simplify the act of content consumption, which is fantastic.

The act of content creation is more fickle and has always required special tools and running statistical analyses falls firmly into content creation. While it is true that most people are not programmers, it is also true that most people are not creating statistical content. Being able to program grants you agility in the face of data analysis that large statistical software packages cannot provide. They move too slowly.

R's core functionality moves fairly slowly as well, but it gives you the tools you need to implement basically any algorithm or any statistical model. This is leading to a lot of innovation by small groups of users, creating packages to fill voids. It feels more like a programming language than a unified piece of software (libraries! command-line!), but this is what makes it flexible.

And if we are being honest with ourselves there is a fundamental fact: point-and-click interfaces do not promote replicability. This might be fine in the private sector, I am not sure. But in the academic world, being able to replicate a finding is crucial.

Posted by Matt Blackwell at 8:27 AM

16 March 2010

Humans make comeback in Zombie models!

Last Halloween, I alerted readers of the social science statistics blog to cutting edge research suggesting that if zombies attacked, humans faced serious risk of extinction.

It turns out that some of these conclusions may have been premature. Some recent research by Blake Messer suggests that if there is terrain that favors humans in some way, then humans may have a better shot at survival.

But it doesn't end there.

Continue reading "Humans make comeback in Zombie models!"

Posted by Richard Nielsen at 12:49 PM

12 March 2010

Putting Statistics in Golf

Here's a neat article in the Wall Street Journal on a new putting statistic recently adopted by the PGA that was developed by researchers at MIT's Sloan School of Management. The article gives a great rundown on the deficiencies of the "putting average" traditionally used to rate pro golfers, then explains in detail how this new statistic improves upon it. Cool stuff!

Posted by John Graves at 12:01 PM