ERIC Number: ED464100
Record Type: RIE
Publication Date: 2001-Nov
A Proposed New "What If Reliability" Analysis for Assessing the Statistical Significance of Bivariate Relationships.
Onwuegbuzie, Anthony J.; Daniel, Larry G.; Roberts, J. Kyle
The purpose of this paper is to illustrate how displaying disattenuated correlation coefficients along with their unadjusted counterparts will allow the reader to assess the impact of unreliability on each bivariate relationship. The paper also demonstrates how a proposed new "what if reliability" analysis can complement the conventional null hypothesis significance test of bivariate relationships. The "what if reliability" procedure helps the researcher determine the extent to which the sample size, as opposed to the effect size, is responsible for the observed statistically significant, or not statistically significant, finding. The analyses illustrate how the sample size needed to detect a statistically significant bivariate relationship decreases as the observed score reliability coefficient pertaining to the independent or dependent measure (theoretically) increases, holding all other factors constant. As such, "what if reliability" analyses will help researchers interpret their results by considering the extent to which the reliability coefficients contribute or fail to contribute to the ability to achieve a statistically significant result. (Contains 3 tables and 28 references.) (SLD)
Publication Type: Reports - Descriptive; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Note: Paper presented at the Annual Meeting of the Mid-South Educational Research Association (30th, Little Rock, AR, November 14-16, 2001).