A Collection of Statistical Myths and Misconceptions Within Psychometrics and Methodology in Education

• In Event: Estimation and Interpretation of Educational Data

Abstract

As the social sciences grapple with issues of replicability and the proper use of statistics, the majority of research has focused on misconceptions and misinterpretations of the p-value. Nevertheless, a large number of statistical heuristics which have permeated our textbooks and classrooms are also unwarranted, yet are seldom questioned. In this session we will show through formal examination and simulation examples that: (1) the Spearman correlation is not a robust estimator of the Pearson correlation; (2) centering in moderated multiple regression can increase multicollinearity; (3) range restriction can increase the correlation and (4) skewness/kurtosis do not always correspond to asymmetry/peakedness. We will show the mathematical conditions where the heuristics hold when possible or examples when it is not.

Authors

• Oscar L. Olvera Astivia, University of Washington

• Edward Kroc, UNIVERSITY OF BRITISH COLUMBIA

• Bruno D. Zumbo, The University of British Columbia