Information Retention as a Function of the Number of Intervals and the Reliability of Continuous Variables

When a researcher constructs a scale, he or she must choose an appropriate number of intervals for the instrument. This study was designed to provide quantitative information to facilitate such decisions. Data were generated based on formulas derived from the classical true-score model. The number of intervals and the reliability of the data were systematically adjusted to determine their effects on information retention. As expected, information retention approached its maximum value when the number of intervals was about eight or more and reliability was near 1.0. Information retention approached its minimum value when the number of intervals was small and reliability was near zero. Moreover, increasing the number of intervals had a greater effect on information retention when reliability was near 1 .O. Tabled information-retention coefficients are presented as an aid to the practitioner who is selecting or creating scales. In addition, suggestions are offered for selecting a useful number of scale intervals.