The Reliability of Sums and Differences of Test Scores

Questions and Answers in the Measurement of Change

“…Despite this obvious and optimal statistical property, authors in the empirical and methodological literatures have criticized the difference score so thoroughly and continuously over the years that investigators have become wary of its use in their research (see, for instance, Bereiter, 1963;Bohrnstedt, 1969;Kes- Linn & Slinde, 1977;O'Connor, 1972). However, recent methodological research has revealed that these deficiencies are perceived rather than actual, imaginary rather than real (Rogosa et al, 1982;Rogosa & Willett, 1983; see also Zimmerman, Brotohusodo, & Williams, 1981;Zimmerman & Williams, 1982a). However, recent methodological research has revealed that these deficiencies are perceived rather than actual, imaginary rather than real (Rogosa et al, 1982;Rogosa & Willett, 1983; see also Zimmerman, Brotohusodo, & Williams, 1981;Zimmerman & Williams, 1982a).…”

Quantitative Methods for Studying Social Context in Multilevels and through Interpersonal Relations

“…Lee and Smith chose to incorporate two measures of achievement (measured in 8th and 10th grades) by defining their outcome as the gain in achievement over the 2-year period (the primary alternative would have been to use 8th-grade achievement as a covariate in a model of lOth-grade achievement). Gain scores, or difference scores, have been much criticized as "unreliable" (e.g., Bereiter, 1963;Linn & Slinde, 1977;see Willett, 1988, for a review); more recently, however, the deficiencies of difference scores have been described as "perceived rather than actual, imaginary rather than reaΓ (Willett, 1988, p. 367; see also Rogosa, Brandt, & Zimowski, 1982;Rogosa & Willett, 1983;Zimmerman, Brotohusodo, & Williams, 1981). The difference score is an unbiased measure of the true change in an outcome, a reasonable measure of growth when one has data from only two time points (Willett, 1988).…”

Chapter 5: Quantitative Methods for Studying Social Context in Multilevels and Through Interpersonal Relations

“…Lee and Smith chose to incorporate two measures of achievement (measured in 8th and 10th grades) by defining their outcome as the gain in achievement over the 2-year period (the primary alternative would have been to use 8th-grade achievement as a covariate in a model of lOth-grade achievement). Gain scores, or difference scores, have been much criticized as "unreliable" (e.g., Bereiter, 1963;Linn & Slinde, 1977; see Willett, 1988, for a review); more recently, however, the deficiencies of difference scores have been described as "perceived rather than actual, imaginary rather than reaΓ (Willett, 1988, p. 367; see also Rogosa, Brandt, & Zimowski, 1982;Rogosa & Willett, 1983;Zimmerman, Brotohusodo, & Williams, 1981). The difference score is an unbiased measure of the true change in an outcome, a reasonable measure of growth when one has data from only two time points (Willett, 1988).…”