|
Abstract:
|
To date, interventions have typically focused on increasing correctness as the primary metric of efficacy across education research. However, correctness alone fails to comprehensively capture the set of competencies displayed by students. Here, we propose a shift in focus to the relative sophistication of problem-solving strategies ordered according to research-based developmental guidelines. We describe a novel modeling approach that treats ordinal strategies as the outcome of interest and explicitly accounts for the differential probability of detecting a strategy for each experimental condition, item, student, and classroom. Our model is estimated using the efficient No-U-Turn Hamiltonian Monte Carlo in Stan. We pilot our analysis on data collected during a kindergarten early geometry intervention situated within an urban school district in a Mountain West US state. We investigate the intervention efficacy of two one-on-one learning approaches relative to a comparison group and describe how to interpret item, student, and classroom random effects.
|
