I wanted to add a brief comment of my own about the exchange between Language Logger Geoff Nunberg and political scientists Tim Groseclose and Jeff Milyo. I'm posting this as a separate item because I wanted to let Groseclose and Milyo speak for themselves, with a simple frame explaining that I was posting the response that they had sent me.
First, here's a bit about the history. Geoff Nunberg posted his critique of the Groseclose and Milyo article here on July 5. Jeff Milyo emailed me on July 21, explaining that he had read Geoff's piece after hearing about it from a Language Log reader, that had written to Geoff (Nunberg) asking if it would be possible to post a response, and that Geoff had suggested that he contact me, since I administer the weblog. I responded that I'd be happy to post their response, and this afternoon Jeff (Milyo) sent it to me. It took some massaging -- he sent a Microsoft Word file, and saving this as html resulted in some pretty strange html code -- but I hope that emacs and I have succeeded in coaxing the output into the form that the authors intended.
Second, I'd like to express my own opinions, such as they are. With respect to the statistical methodology, I don't think I'm in a position to judge. When I first read the Groseclose and Milyo article, my reaction to their "back-of-the-envelope" version was similar to Geoff's. I did realize that the most obvious objections to this version don't apply to the "real" technique that they used. On the other hand, I've noticed that the popular-press discussion of their article has focused mainly on the easier-to-understand "back-of-the-envelope" method, and so it does seem fair to me for Geoff to have criticized it. I believe that I do understand their "real" statistical model, and I look forward to some further discussion about what conclusions its application to in this case licenses. Geoff didn't engage this question, and it seems fair to me for them to complain about this.
With respect to the tone and style of the criticism, it seems to me that there's a certain clash of expectations here. As Geoff pointed out, the Groseclose and Milyo article has been widely discussed in the popular press, where its conclusions have been often presented in a polemical light. Although Geoff can (and I expect will) speak for himself, I took his LL posting to be presented in the rough-and-tumble style of these political polemics, rather than in the typically more subdued style of an academic review. The G&M paper itself somewhat straddles this divide. It's a piece of academic social science, but I get the impression from reading it that its authors also intended to make a political point.
So I feel their pain -- if I had published a scientific article that was criticized with the rhetorical devices that Geoff applied to their paper (starting with "sand sifted statistically is still sand", and moving on from there), I'd be pretty upset too. On the other hand, if I published a political tract that got similar treatment, I'd think to myself "oh good, I'm making enough impact that someone's taking the trouble to attack me", and I'd respond in the same spirit.
Anyhow, I think that it was only fair to give G&M the chance to respond, in the same space, to Geoff's critique.
Posted by Mark Liberman at August 2, 2004 06:20 PM