On Using and Interpreting the Max R Statistic: A Cautionary Note

Abstract

Max r, which is the maximum of all possible split-half reliabilities, has recently been used to reanalyze the psychometric properties of various psychological scales. It is recommended that these revised estimates be interpreted cautiously for two reasons, (1) scale homogeneity is not always the best indicator of reliability and (2) the coefficients cannot be interpreted as equivalent to published reliabilities, which are often based on estimates of the mean split-half reliabilities (e.g., Cronbach's alpha or Kuder-Richardson 20).