A Correlation and Regression Study of the Predictive Validity of the Stanford Achievement Test in the Spencerport Schools

Abstract

The use of achievement tests has become widespread in American education, as they help identify learning gaps or problems for individual students and help school administrators evaluate the effectiveness of the instructional programs in each individual school or district. Given their ubiquity, it is essential to research the reliability and validity of achievement tests in meeting the educational objectives of school districts. This study examines the predictive validity of the commonly used Stanford Achievement Test (SAT) in the Spencerport School District. A sample of 112 students were identified who had spent kindergarten through grade six in the district. The scores of this group’s fourth and sixth grade SATs were subjected to correlation and regression analysis to see how accurately the variation in fourth grade scores explains the variation in sixth grade scores and whether there was a statistically significant difference in the mean performance of male and female students. The author found that for the reading comprehension portion of the test, fourth grade scores explained 53% of the variation in sixth grade scores, while for the mathematics portion, fourth grade scores explained 67% of the variation in sixth grade scores. This means that the reading scores can be interpreted as moderately strong in predictive value while math scores can be interpreted as very strong. The author also found no statistically significant difference between male and female students on either portion of the exam.