Common Methodology Mistakes in Dissertations, Revisited.

N

44

Behavioral Science Research; Case Studies; Doctoral Dissertations; Error Patterns; Higher Education; Information Dissemination; Literature Reviews; Measurement Techniques; Multivariate Analysis; Research Design; Research Methodology; Research Problems; Research Reports; Statistical Significance; Test Interpretation; Test Reliability; Test Use; Writing (Composition)

Dissertations are an important component of the effort to generate knowledge. Thus, dissertation quality may be seen by accreditation and coordinating-board reviewers as a noteworthy reflection on the quality of doctoral programs themselves. The present study reviews methodological errors within Ph.D. dissertations. The illustrative errors are presented within the framework of seven analytic principles, each of which is explored in some detail. These are (1) Because tests are not reliable, but data are, data must generally be evaluated for measurement integrity; (2) Statistical significance testing is of limited importance and should be augmented by other analyses; (3) Multivariate methods are usually vital in behavioral research; (4) Result interpretations should not be based only on standardized weights; (5) Intervally scaled variables should generally not be converted to the nominal level of scale; (6) Covariance corrections are generally either unnecessary or ineffective and should usually be avoided; and (7) Stepwise methods should not be used. Includes six tables, one figure. (Contains 83 references.) (Author/SLD)

1994

ED368771