What I’ve Learned From Cosma Shalizi

Ignore all social scientific results based on statistical inference.  That’s even if you’ve read the underlying research, so that you know you’re not getting snookered by careless popular journalism.

Of course, the three posts I linked to are a bit, ahh…complicated, and I may have oversimplified or overgeneralized them.  Maybe we’ll call my version a layman’s heuristic?

Do I have it right, Cosma?


2 responses to “What I’ve Learned From Cosma Shalizi

  1. I don’t think I would go that far, no. Leaving to one side the big difference between experimental studies (like this cutie) and observational ones, there are certainly worthwhile applications of statistical inference to observational data.
    The work done by Andy Gelman and collaborators on the “red state/blue state” issue, for instance, is valuable and correct, as is, in a very different way, much of the work on the structure of social networks (e.g.). This is off the top of my head on a Sunday morning

    It is true that many of the methods routinely applied are horribly ill-suited to the tasks they are used for. I need to work on a one-line heuristic.

  2. My current take is that my role on these here internets is to:

    1) Be ridiculous
    2) Link to interesting ideas
    3) Provoke people to re-present their insights in a way that I find convenient and congenial

    This has been a success on 1 & 2, and if you ever get that heuristic, it’ll be a success on 3.