UCLA

Statistical literacy is increasingly essential for informed citizenship, yet little is known about how its problem solving skills, Formulate Questions (FQ), Collect Data (CD), Analyze Data (AD), and Interpret Results (IR), develop over time or vary geographically among secondary students. This thesis uses ten years of data (2015-2025) from the LOCUS assessment to examine performance across these four categories among 5,819 AP Statistics students in 11 states and 19 counties, applying fixed-effects regression, hierarchical linear models, random forests, and MANOVA. Results show a robust within-year improvement from first to second semester (10 to 16 percentage points, largest for IR), no consistent linear trend across years after accounting for semester and county, and a COVID-19 disruption reflected mainly in missing second-semester test administrations rather than uniform score declines. County-level clustering explained between 3.2% (FQ) and 14.6% (overall) of variance, with Analyze Data and Interpret Results showing the greatest geographic sensitivity, including a roughly 40-percentage-point gap between Santa Clara and Los Angeles Counties, California, on AD alone. These findings indicate that the four LOCUS categories are separable skills with differing sensitivity to time and place, that Analyze Data remains the most persistently difficult skill for AP Statistics students, and that geographic variation in higher-order statistical reasoning likely reflects broader patterns of educational inequality.