The validity of four measures of change derived from six types of scores

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)


Educational Research

First Committee Member

Richard H. Williams - Committee Chair


The purpose of this study was to examine the validity of four measures of change derived from six types of scores.Validity coefficients were computed for simple gains, Lord's measure, base-free measure and residualized gain. These coefficients were computed for raw scores, percentiles, stanines, normal curve equivalent scores (NCE), normalized z-scores and grade equivalent scores (GES). Convergent validity and discriminant validity were the validities considered. Validity coefficients were found using local and national norms. Data for this study were scores Dade County Public School students received on successive administrations of the Stanford Achievement Test, Intermediate 2 version. Students were tested in the fifth grade in 1987 and in the sixth grade in 1988. The Mathematics Computation score was used as the pretest and posttest measure. Mathematics Applications was used as the criterion for convergent validity. The Reading score was used to compute discriminant validity.Validity coefficients ranged from $-$.030 to.502. GES's produced the largest coefficients in most instances. There was a significant difference between the convergent validity and the discriminant validity coefficients in all situations but one (the two values involved were not significantly different from zero).The choice of norms created significant differences in the convergent validity coefficients for all scores using simple gains and Lord's measure with local norms producing the larger values. Using the base-free and the residualized measures, raw scores had significant differences for choice of norm. GES validity coefficients using local norms were significantly larger than those found using national norms for the base-free measure.Using local norms for all types of measures of change, GES's produced validities significantly larger than those of the other scores. Using national norms, GES's and NCEs had significantly larger values than the other scores for simple gains and Lord's measure (GES $<$ NCE). For base-free and residualized gains, raw scores had validity coefficients significantly less than the other scores.The validity coefficients for all scores conformed to the criteria defined by Williams and Zimmerman. The criteria were extended to include the relationship of the validity coefficient for Lord's measure to those of the other measures.


Education, Tests and Measurements

Link to Full Text


Link to Full Text