The precision of change scores under absolute and relative interpretations

Although some progress has been made in rehabilitating change scores by sho wing that they are not necessarily unreliable, they are still generally tho ught to be less reliable than the measures from which they are derived. How ever, the focus on reliability or generalizability as a measure of the prec ision of change scores may be misguided because both reliability and genera lizability coefficients assume a norm-referenced interpretation, whereas ma ny applications of change scores rely on an absolute interpretation of the change. We examined the precision of change scores in terms of error-tolera nce (E/T) ratios for both relative and absolute interpretations of change s cores. E/T ratios are used to evaluate the error in estimating the change r elative to the tolerance for error in a particular context. Achievement tes t data in which people are crossed with items within categories are used to illustrate the results.