FWD [fort] Re: A Field Guide to Skepticism (3 of 3)

From: Terry W. Colvin (fortean1@mindspring.com)
Date: Mon Jun 09 2003 - 22:54:39 MDT

  • Next message: Terry W. Colvin: "FWD (SK) Re: Find-a-Drug computing project(s) [new subj]"

    --- In fort@yahoogroups.com, "Terry W. Colvin" <fortean1@m...> wrote:
    > Read the book from which that `essay' was extracted, where plenty of
    > evidence is cited. Bear in mind, though, that Dr. Radin's own studies have
    > been assailed from some quarters *within* the `anomalous phenomena'
    > research discipline for their alleged statistical inadequacies. One of
    > Radin's most interesting and provocative recent findings is an effect he
    > calls `presponse', where it is claimed that horrifying images provoke an
    > advance EEG response prior to randomized presentation, compared with
    > psychologically neutral images (as established by standard psych tests).

    Hey, at least he's willing to discuss the matter. I sent him this
    email back in November, after a few previous messages:

    "I am not suggesting any kind of trial-to-trial strategy. Here is
    what I'm wondering:

    Let's assume the subjects are guessing what the stimulus will be, one
    trial at a time, no inter-trial strategy. Some subjects are going to
    do a better job of guessing, purely by chance. When you lump all the
    subjects together, you get no significant results.

    Now you divide the subjects into the consistent and inconsistent
    sets. Is it possible that the chosen method of segregating the
    inconsistent subjects skews the remaining subjects toward a positive
    correlation between guess and stimulus type, thus giving a
    significant result? I think there is.

    Imagine that when subjects guess the stimulus is going to be
    disturbing, but it turns out to be calm, they tend to have an erratic
    response. The subjects with more of these kind of trials, due to
    inferior guessing, will be placed in the inconsistent set. But
    because these trials have a negative correlation between guess and
    stimulus, it will skew the remaining subjects toward a positive
    correlation and a significant result."

    His response:

    "> Now you divide the subjects into the consistent and
    > inconsistent sets. Is it possible that the chosen method of
    > segregating the inconsistent subjects skews the remaining
    > subjects toward a positive correlation between guess and
    > stimulus type, thus giving a significant result? I think there is.

    I see what you mean. And yes, I see how it might introduce an
    artifact.

    I should mention that my current analytical method no longer separates
    subjects by any physiological parameter, before or after the stimulus,
    because of the possibility that such partitioning may introduce
    artifacts (as you've noted). Instead I look at trials for all subjects but
    only the most calm and most emotional targets they saw, where target
    emotionality is pre-assessed by an independent group of judges (i.e., so the
    definitions of "calm" and "emotional" are decoupled from any given subject's
    physiological responses)."

    ~~ Paul

    -- 
    Terry W. Colvin, Sierra Vista, Arizona (USA) < fortean1@mindspring.com >
         Alternate: < fortean1@msn.com >
    Home Page: < http://www.geocities.com/Area51/Stargate/8958/index.html >
    Sites: * Fortean Times * Mystic's Haven * TLCB *
          U.S. Message Text Formatting (USMTF) Program
    ------------
    Member: Thailand-Laos-Cambodia Brotherhood (TLCB) Mailing List
       TLCB Web Site: < http://www.tlc-brotherhood.org >[Vietnam veterans,
    Allies, CIA/NSA, and "steenkeen" contractors are welcome.]
    


    This archive was generated by hypermail 2.1.5 : Mon Jun 09 2003 - 23:06:57 MDT