CLEAR Screener — Data-Driven Assessment
Built from ReadyChecks using Item Response Theory (IRT) and ML validation to measure student preparedness across grades.
Overview
Developed a short diagnostic assessment constructed from a large pool of Carnegie Learning ReadyCheck items. Using Item Response Theory (IRT) and machine learning correlations with external outcomes, the screener was validated to provide reliable, predictive measures of student preparedness.
Approach
- Item Selection & Calibration: Applied IRT models to evaluate item quality and select the best-performing subset.
- Predictive Validation: Correlated screener results with end-of-year assessments to ensure predictive power.
- Data Pipeline: Built robust preprocessing and scoring pipelines for large student datasets.
Impact
The screener enables early identification of at-risk learners with fewer questions, reducing testing time while maintaining reliability. This project bridged psychometrics and ML to create an assessment now used in live educational contexts.
Skills demonstrated: Item Response Theory, ML validation, assessment design, data pipelines.