Identifying Predictors of Students' Skipping Behavior on NAEP Mathematics Assessments: A Random Forest Approach

• In Event: 43.050 - WITHDRAWN: New Research Methods for Large-Scale Assessments

Abstract

Student’s skipping behavior in large-scale assessments such as NAEP, as presented by omitted response, happens when the student is presented an item but decides to skip over it. Such responses are not ignorable under IRT models, and must be filled in with best guesses. However, little empirical evidence exists regarding why students omit responses. Random forest method is used in this study to investigate predictors of omitted responses on the 2015 NAEP mathematics assessment, using variables from the extensive student, teacher, and school questionnaires. A total of 325 background variables besides students’ math achievement are selected and their contribution to the variance of omitted responses were ranked. A random forest based missing data handling method is also implemented and discussed.

Author

• Shenghai Dai, Washington State University - Pullman