« Welcome to New Permanent ELS Blog Editor: Frank Cross | Main | Longitudinal Changes in the Promotion to Partnership Tournament »

31 January 2007


Tracy Lightcap

Gary King has been carrying on a long crusade along the lines Jeremy suggests, with some success. The line of thought stretches all the way back to John Tukey's "Badmandments" in the early '60's.

I wish we could switch over more to resampling stats myself, as I have outlined in comments below. But we might as well admit that this really isn't a scientific controversy; it's a matter of professional boundary maintenance. Knowing and understanding the basics of Fisher-style statistical inference (it does, after all, take at least two tries) is a major rite of passage in the social sciences and the continuation of the "standards" that accompany it are unlikely to be overturned by any controversy over their meaning any time soon. There is simply too much invested by too many people in maintaining the status quo.

Still, it would be intellectually dishonest to avoid questions about our professional practices, albeit that we are unlikely to change them that way. The increasing sophistication of methodological education will put paid to some of the bad practices as time progresses and, as Kuhn points out, the adherents of the older ideas die out.

Jeremy A. Blumenthal

In psychology at least, critics of p-value significance testing have long focused on two "alternatives." The first, confidence intervals, might be thought of as in the same family; the second, though - effect size (r, d, phi, etc.) - is less dependent on sample size and gives, in my opinion, a better sense of the strength of a finding. It also, arguably, can give a better sense of a finding's "importance," though I recognize the loaded and variable nature of that term.

The comments to this entry are closed.


January 2015

Sun Mon Tue Wed Thu Fri Sat
        1 2 3
4 5 6 7 8 9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31

Site Meter

Creative Commons License

  • Creative Commons License
Blog powered by Typepad