Alex Tabarrok has just had a co-authored paper published that uses a simulation-based approach (using simulated “placebo laws”) to help test whether the effects of certain types of dummy (binary) variables in a time-series are statistically significant. It seems like a fairly interesting approach, which I’ll have to bear in mind next time I do any time-series stuff (most of my data tends to be cross-sectional, however).

And, the substantive topic of Alex’s paper will no doubt be of interest to those who want to fight over John Lott’s More Guns, Less Crime.

Will Baude ponders the procedure of assigning students to discussion sections. My gut reaction is: “don’t have discussion sections”—none of the political science courses I ever took had them, and I’m not really all that sure they add much value for anyone involved in the process. I suppose they provide a way to get grad students some teaching experience to slap on the vita without having to give them the responsibility of teaching a real course and running the risk that their first outing will be an unmitigated disaster.

Such things do happen, mind you, but I’d rather grad students crash and burn during a lecture early in their careers rather than emerging from school with zero experience besides asking ten undergrads to discuss their feelings about Plato’s allegory of the cave*—and then crashing and burning repeatedly on their way to being unceremoniously canned at their third-year review.

By the way, my recommended procedure is to have the discussion sections listed separately in the schedule of classes (or, as a bookkeeping exercise, each lecture-discussion combo is considered a separate course even though the lectures all meet at the same place), so students sign up for them directly. This neatly avoids the issue of schedule conflicts since, if the scheduling program is doing its job, there won’t be any.

Michael Jennings discusses the slow but steady progress of LEDs in replacing traditional incandescent lights. As he notes, they’ve become particularly common in traffic signals because they are brighter, last much longer, and have significantly lower power consumption than traditional lights.

Stephen Green ponders whether Howard Dean’s candidacy is stagnating in the face of surges from Wesley Clark (in New Hampshire, as he’s had the whole state to himself while the rest of the Dems pander to Iowans prior to next Tuesday’s essentially meaningless precinct caucuses) and John Edwards (who’s picking up endorsements and favorable media coverage in Iowa).

At this point, the narrative for Iowa is pretty much written:* Edwards surges to a surprisingly strong third-place finish, and Dick Gephardt fails to live up to expectations in his own back yard against Dean, effectively starting the “death bells” for Gephardt’s campaign—with the nails to the coffin coming when he finishes spectacularly poorly in New Hampshire.

So, what’s the New Hampshire narrative? Today’s polls still show Dean with a statistically-significant, but rapidly eroding, lead over Clark. If Dean and Clark finish within single digits of each other, Dean fails to live up to expectations—and has to hope that Clark, Edwards, and Gephardt divide the South Carolina electorate enough for Dean to finish #2 behind Edwards. If, on the other hand, Dean gets a double-digit win over Clark in New Hampshire, that’s probably enough to make him the designated frontrunner and tip the balance in the non-S.C. February 3rd primaries through favorable media.

Stay tuned, things are about to get interesting…

* Update: Ok, maybe not… where the heck did Kerry come from?