Technical Brief on the 1999 Statistical Model
Factors Associated with Student Achievement
Prior achievement is the most powerful predictor of school success, accounting for about 50% of the variation in student achievement scores. Barring unforeseen circumstances, a child who is achieving academically will probably continue to achieve. This means that no matter what the family's circumstances, any child who consistently arrives at school adequately nourished, rested, healthy, and feeling safe and stable will be able to acquire and accumulate learning. This is what we mean by "ready to learn." A child's readiness to learn, especially in the early years, positions that child for lifelong success with learning.
A wealth of studies shows that family background characteristics are closely related to student achievement. Schools with less economically privileged students, for example, almost always have lower achievement scores. (When RI rank orders their state test results, the results closely mirror the socioeconomic status (SES) of the district; thus, high income districts have high scores and scores drop with a strong correspondence to the relative drop in income.) Changes in many other characteristics (variables) have also been shown by many research studies to correlate closely to student achievement. These include:
Prior achievement on aptitude»Participation in free and reduced lunch program
Additionally, at higher levels of the system beyond groups of individual students, we know that a number of other factors are associated with student achievement such as school settings (urban, rural, suburban), per pupil expenditure, policies and practices within schools or school districts, and community characteristics (e.g., job market, tax support).2
Performance indicators of school effects have been systematically collected in a variety of places in the U.S. and elsewhere.3 Attempts to identify effective schools have created many controversies over the kinds of data to be collected, the appropriate methodologies to be employed and the interpretation of specific results. The first few years of Information Works! will probably witness similar controversies. The researchers who constructed the 1999 RI model are not wedded to it. This year's model was based on various data sources that were already available and took into account the strengths and weaknesses associated with each available data set. As both the quality of the data improves and new research is accomplished, the model will evolve and become increasingly sophisticated.
The following few sections of this brief describe some general statistical principles that are important for understanding the statistical model used in this year's Information Works! These principles are then applied specifically to the model that was created to generate the second field of the Rhode Island school and district reports. Readers already familiar with hierarchical regression analysis may wish to skip directly to the sections on Multiple Regression and the RI Model.