Skip to content
 

A whole new kind of z-statistic

Here.

17 Comments

  1. elissa says:

    Aaaaaaaaaaaaaaaaaaaaaand win.

  2. FH says:

    This couldn't wait until 4/1/2010 lest you get scooped? Or preparing the ground for results of a Leprechaun survey taken today?

  3. Richard D. Morey says:

    "Zombies are believed to have very low rates of telephone usage and in any case may be reluctant to identify themselves as such to a researcher." I don't know, every time I call Bank of America's customer service line, I get a zombie.

  4. zbicyclist says:

    This is very timely. Around here, I've successfully introduced the term "zombie panelist" to refer to longitudinal respondents who seem to have dropped out of sample, but later return.

    This leads to sentences in reports like this one from September 2009: "The zombies are a de minimis nuisance in processing an annual sample by month since it's only the annual boundary we need to worry about; they are going to be more serious when we are doing weekly maintenance."

    And this email from a statistician at another firm: "In January, 2008 136 zombie households were added."

    Were it not for research niceties like "make sure your operational definitions are compatible" I could provide you with several hundred zombies for study.

  5. Zubon says:

    The referee comments include requests for more analysis on the graphs and for braaaaaaains.

  6. Andrew Gelman says:

    All I can say is, I tried to post it on Arxiv but they wouldn't take it. Which is pretty ridiculous given that they accepted the paper linked to here, and all of its jokes were unintentional, I believe!

  7. Steve says:

    But what about philosophical zombies?

  8. Anne says:

    You could try vixra.org, for those papers too "controversial" to be accepted on arxiv.org. Or, at a higher tone, you could aim for Rejecta Mathematica, though I suppose you'd have to have it rejected from some other journal first.

  9. Phil says:

    It's nice, but I do suggest some edits. My main comment is that you should play it straight.

    –P

  10. Naadir Jeewa says:

    This paper is seriously flawed. Needs more lolcats.

  11. Manolo Romero says:

    Good Ol'uncle George was published and cited before me (not my real uncle btw). I found hilarious the mention of a cite called ZDate… in "Canadian" it would read "zed-date".

  12. Kevin Denny says:

    You really need to re-cast this problem in a Bayesian framework & then its all trivial. I thought you'd know that.

  13. Terrell says:

    Well played.

    For serious science, at least spell it LaTeX.

  14. Andrew Gelman says:

    Manolo: I'm glad somebody caught that ZDate line. When writing things, I put in all sorts of gags, and I'm never sure whether they get noticed.

    Kevin: But our method is Bayesian; check the references. The inference might be trivial, but the data collection and model building steps are not trivial at all.

    Naadir: Yes, true, but "Needs more lolcats" is a valid criticism of any article. No matter how many lolcats a paper might have, it can always be improved by adding more.

    Anne: I took a look at vixra.org, and boy were those papers bad! Until I had this difficulty with the zombies paper, I always thought Arxiv would accept anything. I still think they accept basically anything. An archive of papers rejected from Arxiv is gonna be pretty bad. Much worse than the comments here, for a start.

  15. zbicyclist says:

    Re figure 2: Peaks here seem to precede zombie attack peaks. Coincidence or causality?

    http://www.google.com/trends?q=gelman

  16. Bobmeister says:

    The paper is interesting, but contains two phatal phlaws. You say that one person knows, on average (perhaps mean?) about 750 people, thus a survey of 1500 people can give us indirect information on about a million people.
    Flough # 1. What makes you assume that all Anmericans are people? It is well known that all members of the opposite party are non-persons.
    Pflogle # 2. Who knows how many of the sample know a lot of the SAME PEOPLE (or sane persons). That will knock a hole in your calculations, won't it?

    Hah!

    With fondest regards.

  17. Andrew Gelman says:

    Bob:

    Regarding your point #1, there's some evidence for political polarization in social networks (see figure 8.12 of Red State, Blue State).

    Regarding your point #2: There are 300 billion Americans, so only a very small percentage of the 750*1500 people in the indirect sample will overlap.