PSY475 dqs INTRODUCTION PSYCHOLOGICAL MEASURMENT SECOND EDITION

profilemasterazjie

 1    The following link (http://prefrontal.org/files/posters/Bennett-Salmon-2009.pdf) is a great conference poster that reports the results of an FMRI study of a dead Atlantic salmon.  Although FMRI is a very sophisticated research tool, the statistics used to analyze FMRI data haven't always been as sophisticated as they should be.  This poster highlights a very basic (and entirely avoidable) statistical problem with running too many statistical comparisons.  Can anyone identify what this problem is and describe a way to avoid it?    200 words and APA FORMAT

 

    2  In the realm of ability and achievement tests, most selected-response items are simply scored correct/incorrect; one point is awarded for a correct response, zero points for an incorrect response.  Then, the score on the test is the total number of correct responses. However, there are other ways to score these items. One variation involves awarding partial credit for selecting an option that is not the best possible answer and yet is not clearly wrong. Another variation involves giving extra weight to items that are especially important. For example, for purposes of computing the total score on the test, a correct response to certain items might be worth three points, to other items two points, and for the remaining items one point. Still another variation in scoring items for which there is a correct answer involves application of the correction for guessing (see Chapter 3, pp. 74-75). Comparisons of the simple 0/1 scoring method with these more complicated methods have been the subject of much research.  The question is whether the more complicated scoring systems yield more reliable and/or more valid scores than the simpler 0/1 scoring. The answer comes in two parts. First, the more complicated systems usually yield only slightly better (more reliable or more valid) scores. On the second part of the answer, authors differ. Some say that with such marginal gains, it is not worth bothering with the more complicated systems (see, e.g., Nunnally & Bernstein, 1993). Others say that with the ease of modern computer scoring, the so-called complications are trivial, so any gain in reliability or validity is worthwhile. For description of some of these item- and option-weighting procedures, see Millman and Greene (1993). (Hogan, 2007, p. 214)"

 

Given this information, what do you think about using very complicated scoring methods to gain a little additional reliability and validity?  Is the loss of simplicity worth the small gains in reliability and validity? ( 200 WORDS)

 

    • 9 years ago
    • 5
    Answer(1)

    Purchase the answer to view it

    blurred-text
    • attachment
      statistical_problems.docx
    Bids(1)