Karen Rosenbloom posted:
> I am asking for your advice. I have recently conducted a survey of
> attitudes toward research from a professional group. There are some
> outliers (+/- 3SD) that I would eliminate , but others conducting the
> research with me feel that this might be a minority view, and should not
> be eliminate from the dataset......any views or references that I should
> read to confirm my view, or thiers?
Attitudes commonly are surveyed by Likert-type scales. Although the original
method described taking averages & such on arbitrary numeric scale numbers,
others have noted that different response combinations all can lead to the
same summary score and that different scales produce different results.
Thus, reporting the frequency distribution (including "outliers" unless it
can be shown that they represent individuals who filled out forms
incorrectly, or shouldn't have been included in the sample, etc.) rather
than means and standard deviations would seem more appropriate. In short, I
don't think we can remove "outliers" arbitrarily just because they're beyond
3 standard deviations and also question the validity of summarizing this
descriptive data through means and standard deviations (vs. bin counts, bar
charts, histograms, box plots, etc.).
Business or market-oriented research methods texts might be helpful. See,
for example, MARKETING RESEARCH METHODOLOGICAL FUNDATIONS, 7th ed.
(Churchill GA Jr, Dryden Press, TX, 1999) or BUSINESS RESEARCH METHODS, 6th
ed. (Zikmund WG, Dryden Press, TX, 2000).
David Birnbaum, PhD, MPH
Clinical Assistant Professor
Dept. of Health Care & Epidemiology
University of British Columbia, Canada
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|