• 9 years - 140 posts - Latest - RSS
• OP: 6 Goods vs 5 No Goods
• Thread: 795 Goods vs 639 No Goods

1. Columban
   

   Mike,

   I read your paper. Comments to help you out:

   1) identification and estimation are separate things. And matching helps with model dependence only for estimation. Comparing results when the conditioning set changes is about identification and there is no reason to think that moves across different identification assumptions will be smooth. You confuse this in your paper, and if I as a little grad student did that, I would be savaged and would have failed my qual exam.

   2) be more careful about finite sample versus asymptotic issues with regard to different matching methods.

   3) data mining: see Rubin's design versus analysis article. Matching methods have the feature that one can set them up without any outcome data.

   You made yourself look bad. But you seem like a smart guy, and I'm sure you will do better in the future.

   To PSR: why are we discussing the worst paper at polmeth instead of the good ones?

   9 years ago # QUOTE 14 YEA 11 NAY!

2. America
   

   To PSR: why are we discussing the worst paper at polmeth instead of the good ones?

   You read every paper and can definitively say that this was THE worst? Or you're engaging in a bit of good ol' anonymous hyperbole?

   9 years ago # QUOTE 4 YEA 7 NAY!

3. Cherice
   

   I am confused. Why does a presentation need to lead to people's heads exploding and people laughing at others? Is this elementary school? Seriously you people are deluded. Don't you realize that no-one reads or cares about anything that you do? It makes literally ZERO difference to anything in the world that matters outside your small circle of math nerds. Go get a mojito and chill out for god's sake.

   BECAUSE POLMETH

   9 years ago # QUOTE 3 YEA 1 NAY!

4. Merton
   

   So now we know that Mike Miller is probably one of the biggest trolls on PSR. Nice to know...

   9 years ago # QUOTE 14 YEA 9 NAY!

5. Lunete
   

   Columban, I genuinely thank you for actually responding to me substantively. Responses:

   1) The fact that matching reduces model dependence only for the parametric estimator is precisely my point! However, estimates are a product of both the matching and estimation stages. Keep in mind that Ho et al. similarly change the conditioning variables -- I'm simply extending that to the matching stage. Labeling that `identification' doesn't make it suddenly cease to matter for your estimate.

   2) The paper doesn't say anything about asymptopia. All actual data is finite, so I'm concerned with that.

   3) If someone is going to data-mine (deliberately or not), is the half-second extra step really going to stop them? After all, I can set up my regression equation without reference to data, too -- the problem is that I can always go back to the well when I don't like the finding. That's true of matching, too.

   I owe you a beer, thanks!

   ^Ha, to Merton, I didn't even know this existed until someone e-mailed me. I'm stuck in a hotel for the night, so I have nothing better to do. May as well get some paper feedback!

   9 years ago # QUOTE 13 YEA 14 NAY!

6. Columban
   

   Mike,

   You are missing the points made. Take the data mining issue. The propensity score is an ancillary statistic, a model of Y is not. This is a fundamental and important difference. If you want to argue that we should think about ancillary statistics differently, cool. Then you will have to discuss the work of Cox, Cochran and many others. That could be an interesting paper. But what you have written and said makes some believe, rightly or wrongly, that you don't even know what an ancillary statistic is let alone give any reason for why we should think about them differently. Similar statistical issues arise with the other points made.

   9 years ago # QUOTE 17 YEA 9 NAY!

7. Cherice
   

   Dude is trying to make a good point, and ya'll are just being a bunch of picky, terminology obsessed bitches instead of trying to understand. Do you get that this bulls**t is why the rest of the discipline thinks you guys are assholes? (and why other fields think you're morons?)

   9 years ago # QUOTE 14 YEA 17 NAY!

8. Cyrille
   

   Columban: Just stop trying to help Mike Miller. He is your basic socially awkward guy who doesn't really listen to people trying to help him. I should know, I went to grad school with him. He really just doesn't know how to listen, which I think you're finding out right here on this thread.

   9 years ago # QUOTE 15 YEA 14 NAY!

9. Staci
   

   wow. are you academics or kindergartners? science is a collaborative project: help each other out. this pettiness doesn't advance knowledge for anyone.

   9 years ago # QUOTE 10 YEA 7 NAY!

10. Stephane
    

    ^Mike (because I know that's your post), the fact that you're turning to PSR for help on your so-called "methods" paper is the least of your problems...

    9 years ago # QUOTE 11 YEA 9 NAY!

11. Staci
    

    of course! you've figured me out...!!! because, surely, mike should be the only one interested in fostering collegiality among soc. scientists.

    give me a break.

    no... i'm just an observer who thinks it's sad that people think it's acceptable to treat each other this way. i'm also interested in how this affects science practices. but whatever. feel self-fulfilled in your cynicism.

    9 years ago # QUOTE 5 YEA 0 NAY!

12. Xenia
    

    God, you people are horrible.

    9 years ago # QUOTE 8 YEA 0 NAY!

13. Lunete
    

    Columban,

    Sorry, I'm not following your point. Much like Ho et al., I'm interested in the practical issue of how much estimates shift around and how often you can data-mine (i.e., manufacture a significant treatment when its effect is really 0). I show that it's easier when using matching (with either PSM or CEM). What does the propensity score being an ancillary statistic have to do with that?

    9 years ago # QUOTE 6 YEA 6 NAY!

14. Josepha
    

    Why is there so much more interest in Jr. Profs attacking each other constantly (in a very junior high way), than the senior people who actually have power and good jobs? Is it really like Logan's Run with you people? Once you're 30 you're dead or of no interest?

    Up-thread there was mention of John Londregan. Only 10-12 years ago he was La Crème de la Methods. Now he seems to be in the Witness Protection Program, somewhere in Mercer County. No one really knows what became of him, even though he could not have a better job. Maybe people seem him as the Ghost of Christmas Future and are trying to look away? Or maybe he's a happy guy who plays golf and consults. I don't know, but surely someone does. Happy or not, I guess it's a reminder of the fact that methods people tend to burn out really young, while substance people can keep going.

    9 years ago # QUOTE 8 YEA 3 NAY!

15. America
    

    Josepha:
    Why is there so much more interest in Jr. Profs attacking each other constantly (in a very junior high way), than the senior people who actually have power and good jobs?

    This question has been asked many times, but the answer seems pretty obvious: once you're established you're pretty much settled into your role, for better or for worse. But the ABDs and fresh APs are all jockeying for position, deluding themselves about their potential, and striking out at all potential obstacles to their deluded sense of self-worth. That explains the psychology behind many of the threads and recurring themes on this board, not just the p!ss!ng contests.

    9 years ago # QUOTE 2 YEA 1 NAY!

16. Virgil
    

    Josepha:
    Happy or not, I guess it's a reminder of the fact that methods people tend to burn out really young, while substance people can keep going.

    Yet Londregan's best work is substantive, on legislative organization in latin america.

    9 years ago # QUOTE 3 YEA 2 NAY!

17. Josepha
    

    ^
    You don't get to run away from this guy. JBL won the first Pol Meth prize back in prehistoric days for an article in Political Analysis:
    http://h-net.msu.edu/cgi-bin/logbrowse.pl?trx=vx&list=h-polmeth&month=0009&week=b&msg=1AFakRhWc8M3x039MI64nQ&user=&pw=

    He's a "Fellow of the Society for Political Methodology" from the first cohort:
    http://polmeth.wustl.edu/about.php?page=fellows

    He's a methods OG and there's no way around it. Saying he's really a Latin Americanist (or a legislative scholar or whatever) is like the conservatives who decided GW Bush "wasn't really a conservative" (what a big spender!) around when his poll numbers started to tank.

    Anyway, the point wasn't to attack him (as if he'd care), but just to ask what happened to him. If you redefine him as something else there are lots of other examples to prove the methodologists burn out point, but I was just curious about his story.

    But if you want the twenty-something cage match, you're welcome to it.

    9 years ago # QUOTE 4 YEA 0 NAY!

18. Linton
    

    OK Polmeth folks. Keep this thread going please. The rest of the us are enjoying watching what a sad bunch of folk you are. It's definitely a win for the discipline as a whole.

    9 years ago # QUOTE 4 YEA 2 NAY!

19. Olivier
    

    Bump.

    Is the program this year out yet? In these turbulent times, Polmeth is truly my source of peace and optimism. Thanks, Polmeth!

    5 years ago # QUOTE 1 YEA 1 NAY!

20. Olivier
    

    You guy suck so bad. I was just asking for the Polmeth program and you guys keep talking about these two strangers and their pubs. This is why academia sucks so bad.

    5 years ago # QUOTE 8 YEA 1 NAY!

Reply »