Abstract
This study evaluates the educational effectiveness of short-cycle 2D/3D modeling in elementary school as a tool for developing mechanical, technical, and spatial thinking. By integrating statistical rigor with practical interpretability, we bridge the gap between research findings and classroom applicability: beyond p-values, we report effect sizes (Hedges’ g, r) and the probability of superiority (CLES) to quantify pedagogical impact. Scientific novelty lies in the systematic application of the g–r–CLES triad for educational data interpretation, emphasizing probabilistic insights into learning gains. Practical significance is demonstrated through a compact, technology-accessible intervention format designed for seamless integration into standard curricula. Methodology: A parallel-group experiment (Experimental/Control; grades 1–4, N = 172) employed a pre-post design using the Bennett test. Statistical analysis (Statistica 10) included: Wilcoxon tests for within-group shifts; Mann–Whitney tests for between-group differences in gains (Δ = Post − Pre); Reporting of p, Z, r = Z/√N, Hedges’ g, and CLES (derived from U and d under normal approximation). Results: Experimental groups showed large pre-post improvements (p < .001; r ≈ 0.85–0.88; g ≈ 1.9–2.4; CLES ≈ 0.91–0.95); Control vs. experimental comparisons consistently favored the intervention (CLES ≈ 0.76–0.86), with an aggregate effect of Z = 2.83 (p = .004), r = 0.42, g ≈ 0.96, and CLES ≈ 0.74–0.75. Practical interpretation: In ~75% of cases, a student in the intervention group outperformed a control peer in learning gains. Contributions: Evidence that brief 2D/3D design cycles yield statistically robust and pedagogically meaningful effects; A replicable analysis framework combining classical effect sizes with probabilistic benchmarks; Implementation guidelines for classroom modules, including growth-threshold monitoring. Key words: STEAM, elementary school, 2D/3D modeling, Bennett test, Hedges g, CLES.