Empirical Evaluation of Formulae for Correction of Item-Total Point-Biserial Correlations
Ronald A. Berk · Educational and Psychological Measurement · 1978
Three formulae developed to correct item-total correlations for spuriousness were evaluated. Relationships among corrected, uncorrected, and item-remainder correlations were determined by computing sets of mean, minimum, and maximum deviation coefficients and Spearman rank correlations for nine test lengths. The results indicated that the Guilford or Cureton formula should be applied to item-total correlations below .4 in item pools with as few as 18 to 20 items.