In a correlation meta-analysis across 69 course offerings, diffusion-based centralities correlated with academic achievement with high combined coefficients and predictive intervals¶

Claim · All claims
Evidence · 1 study · quant-synthesis r2 · q3
Currency · newest study 2021 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Random-effects meta-analysis pooling correlations of centralities with final grades across 69 courses showed diffusion-based centralities had the highest and most consistent combined correlation coefficients and highest predictive intervals among the measures tested. → Saqr 2021

Evidence¶

Saqr 2021¶

Saqr, M., & López-Pernas, S. (2021). Modelling diffusion in computer-supported collaborative learning: a large scale learning analytics study. Intern. J. Comput.-Support. Collab. Learn. https://doi.org/10.1007/s11412-021-09356-4

q3 · i? · quant-synthesis · r2

Correlation meta-analysis (17 meta-analyses, one per centrality measure) pooling Fisher's z transformed correlations across all 69 course offerings by inverse-variance weighting under a random-effects model. The article reports diffusion centralities "had the highest and most consistent combined correlation coefficients" and the highest predictive intervals.

"The results of the meta-analysis showed that diffusion-based centralities had the highest and most consistent combined correlation coefficients with academic achievement as well as the highest predictive intervals"

Discussion¶

Related Claims¶

• A worked network example shows diffusion-based centralities capturing uptake and argumentation more accurately than traditional centrality measures
• Multigraph indegree centrality is consistently and positively correlated with student performance across 12 PBL courses, with stronger correlations than other configurations in eight courses
• Multigraph outdegree centrality is the most robust participation indicator, with significant positive correlations in all courses, while weighted outdegree was negatively correlated in two courses
• Diffusion-based centralities show moderate-to-strong statistically significant positive correlations with the frequency of information sharing and argumentation utterances in CSCL
• Multigraph closeness centrality is the most consistent of the three configurations for correlating closeness with performance, though less robust than degree and eigenvector centralities — a narrower finding that bears on this claim
• Betweenness centrality shows the least consistent correlation with performance, being largely insignificant or negative across configurations except in one course — related
• Student roles discovered by clustering diffusion-based centralities were validated with coded data and relate meaningfully to academic performance — related
• Diffusion-based centralities elucidate information exchange and uptake in large-scale collaboration, validated against achievement and qualitative analysis, according to the editors' account of Saqr and López-Pernas's study — a broader claim this one bears on