Re: "Statistically demonstrable"

From: Eric Bohlman (ebohlman_at_omsdev.com)
Date: 07/02/04

  • Next message: Jerry Dallal: "Re: "Statistically demonstrable"" 
    Date: 2 Jul 2004 23:53:07 GMT

    Jerry Dallal <gdallal@nospam.tufts.edu> wrote in
    news:DmWEc.69$ZK5.2001@news.tufts.edu:

    > Professor Rubin's comments show that "statistically demonstrable" is
    > not the panacea I first thought it was. However, I'm not yet
    > convinced that it's worse than "statistically significant". I'm
    > curious to read others' views on the subject as well as hear other
    > proposals to avoid the perils of "significant".

    The problem with "demonstrable" is that, like "significant" it's ambiguous
    (as most words are), with one of the ambiguous senses having connotations
    that you're specifically trying to avoid. How about "statistically
    noticeable" or "statistically observable"? Here the connotation is that
    the results *appear* to be distinguishable from the "background noise," but
    you aren't implying how much of a difference there is, and leaving open the
    possibility that you're really seeing a "spike" in the background noise.

    Or maybe "detectable" in the sense that an effect, if there's any, can be
    either below or above the "detection limit" of a particular methodology.

  • Next message: Jerry Dallal: "Re: "Statistically demonstrable""