Predicting Public School Students at Risk for Standardized Testing Failure

Jennifer Brite, City University of New York (CUNY)
Christina Pollari, CUNY School of Public Health
Jasmine Abdelnabi, CUNY School of Public Health
Ragheed Al-dulaimi, CUNY School of Public Health
William Stumbo, Xerox
Levi Waldron, CUNY School of Public Health

Although standardized testing is taking a more central role in school districts nationwide, models to predict performance are lacking, even as more data is gathered about students. This study will use data from Massachusetts Comprehensive Assessment System, a standardized test taken by all public school 10th graders in Massachusetts. We will use principal component analysis to group demographic, personality trait, academic courses, and career plans variables taken from a student questionnaire and other sources. We will use ROC analysis to identify the components with the highest predictive power, and will then build a model using 2013 data. Finally, we will evaluate our model by using it to predict scores in previous years.