Beyond the Usual Suspects: A Random Forest Analysis of Precollege Predictors

• In Event: AERA Roundtable Session Thursday 3:35 pm Four Seasons Ballroom 2-3
  In Roundtable Session: From Potential to Performance: Understanding Drivers of Student Success (Table 12)

Abstract

Higher education research employs precollege predictors when examining collegiate experiences and success outcomes. The present study explored expanding these predictors to include characteristics about students’ home communities and high schools, representing a shift to include structural variables that may account for how systems of oppression shape students’ lives. Using a multi-institutional sample of 5,199 lower-income students, we employed a random forest algorithm to compare the relative importance of 22 predictor variables across 39 collegiate outcomes. Results reinforce high school GPA as a crucial predictor, with race/ethnicity predictors varying considerably in relative importance. Community-level and high school variables tend to be as important as individual attributes, which suggests that these contextual measures should not be overlooked in future analyses.

Authors

• N Tennessen, University of Iowa

• Nicholas A. Bowman, University of Iowa

• Shinji Katsumoto, University of Southern Mississippi