Abstract

Background
Statistical literacy and quantitative reasoning have become essential competencies across diverse academic disciplines. However, statistics courses often face challenges, including negative student attitudes, high anxiety levels, and suboptimal learning outcomes. Traditional lecture-based instruction has proven insufficient in addressing these challenges, prompting a shift toward more interactive pedagogical approaches. Cooperative learning (CL), endorsed by the American Statistical Association in their Guidelines for Assessment and Instruction in Statistics Education (GAISE), has emerged as a promising alternative. Despite widespread adoption, empirical evidence regarding CL's effectiveness in statistics education remains inconsistent, with studies reporting mixed results.
Purpose
This meta-analysis aims to quantify the overall effect of cooperative learning strategies on student performance in college-level statistics courses. By synthesizing evidence from multiple studies, we seek to resolve discrepancies in the existing literature and provide educators and policymakers with data-driven insights for optimizing instructional practices in statistics education.
Methods
We conducted a comprehensive systematic review and meta-analysis of studies examining cooperative learning interventions in college statistics courses. The analysis included 14 studies, encompassing 2,035 participants. Eligible studies employed quasi-experimental or experimental designs comparing cooperative learning approaches with traditional instruction methods. Effect sizes were calculated using standardized mean differences (Cohen's d). We also conducted moderator analyses to examine the influence of class size and academic discipline on CL effectiveness. Sensitivity analyses using leave-one-out methods were performed to assess the robustness of findings.
Results
The meta-analysis revealed a significant moderate positive effect of cooperative learning on student performance in statistics courses. The random-effects model yielded an overall effect size of Cohen's d = 0.67 (95% CI: 0.50-0.85, p < 0.001), indicating that students in CL environments substantially outperformed those receiving traditional instruction. The fixed-effects model showed a slightly larger effect (d = 0.78, p < 0.001), though substantial heterogeneity was detected (I² = 67.82%, Q = 37.29, p = 0.0002), suggesting variability across studies due to contextual factors.
Positive effects were consistently observed across different class sizes and academic disciplines, though effect sizes varied. Both STEM and non-STEM courses benefited from CL interventions, demonstrating the approach's broad applicability. Sensitivity analyses confirmed the stability of results, with no single study disproportionately influencing the overall conclusions.
Conclusions and Implications
This meta-analysis provides compelling evidence supporting the effectiveness of cooperative learning in enhancing student performance in college statistics courses. The moderate-to-large effect size suggests that CL strategies can meaningfully improve statistical learning outcomes while potentially addressing common barriers such as statistics anxiety and negative attitudes toward the subject.
The findings have important implications for statistics education: (1) educators should consider integrating CL methods into their teaching practices, (2) institutions should provide professional development opportunities for instructors to implement CL effectively, and (3) curriculum designers should incorporate CL principles into statistics course structures.