skip to main content skip to footer

The Contribution of Clusters of Biographical Data to Prediction

Connolly, John A.
Publication Year:
Report Number:
ETS Research Bulletin
Document Type:
Page Count:
Subject/Key Words:
Background, Biographical Inventories, Longitudinal Studies, Performance Factors, Predictive Measurement, A Study of Academic Prediction and Growth


Previous studies indicated that biographical items predict test scores. This study was concerned with the individual and joint predictive effectiveness of homogeneous scales of biographical data. Specifically, three questions were asked: 1. What dimensions of personal background predict performance on aptitude and achievement measures? 2. Do personal background cluster scores add to the level of prediction of final performance that can be obtained from initial test scores on the same test variable? 3. Do personal background cluster scores add to the accuracy of prediction that can be obtained from a comprehensive battery of initial cognitive measures? The 169 items of the Growth Study seventh-grade BEQ were cluster analyzed, separately for boys and girls, to develop homogeneous scales to be used in prediction. A sample of 300 males and a sample of 300 females were selected from those students who entered the Growth Study as fifth graders in 1961 and who were tested again in 1963 and 1965. A stepwise multiple regression analysis was performed using their STEP, SCAT, and BEQ cluster scores. The cluster scores gave a moderately high prediction of subsequent test performance when used alone. In combination with measures of initial status, the scores added small, though statistically significant, increases to prediction. Used with a battery of earlier test measures, BEQ clusters added almost nothing to the prediction.

Read More