In pairwise WD analyses, learner trajectory similarity dominates variation in percentage-overlap values (57.90% of variation) while remaining design factors are essentially zero¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2 · i3 large
Currency · newest study 2012 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i3 In the five-way ANOVA of percentage-overlap WD values, differences in learner trajectories dominated all other sources of variation (57.90% of variation), with a decreasing trend in mean percentage-overlap values across the three trajectory-similarity groups and all remaining effect sizes essentially zero. → Sweet 2012 q2 i? In design matrix 1 games, five of the six prototypical learners—those who make actual learning gains throughout the game—remain rather well separated by WD pseudo-confidence bands after a few evidentiary segments. → Sweet 2012 (2)

Evidence¶

Sweet 2012¶

Sweet, S. J., & Rupp, A. A. (2012). Using the ECD Framework to Support Evidentiary Reasoning in the Context of a Simulation Study for Detecting Learner Differences in Epistemic Games. Journal of Educational Data Mining, Special Issue, Article 5, Volume 4. doi:10.5281/zenodo.3554649

q2 · i3 · design · r2

Five-way ANOVA on percentage-overlap WD values (trajectory-similarity group, task difficulty, specificity, content, complexity; 252 cells) showed "differences in learner trajectories dominated all other sources of variation" with remaining effect sizes essentially zero.

"The results showed that differences in learner trajectories dominated all other sources of variation ( 2 = 57.90 %) with a decreasing trend in mean percentage-overlap values across the three groups; all of the remaining effect size values were essentially zero."

Sweet 2012 (2)¶

Sweet, S. J., & Rupp, A. A. (2012). Using the ECD Framework to Support Evidentiary Reasoning in the Context of a Simulation Study for Detecting Learner Differences in Epistemic Games. Journal of Educational Data Mining, Special Issue, Article 5, Volume 4. doi:10.5281/zenodo.3554649

q2 · i? · design · r2

Empirical 95% pseudo-confidence bands for WD (Figure 10, design matrix 1, medians across 100 replications) show that five of six learner types "remain rather well separated after a few evidentiary segments / tasks", while band width increases slightly across segments.

"five out of the six learners – the ones who make actual learning gains throughout the game (i.e. the ones who do not have a flat trajectory) – remain rather well separated after a few evidentiary segments / tasks"

Discussion¶

Related Claims¶

• Easy, highly specific simulated tasks compress the range of the WD statistic across learner types, while well-designed tasks widen it — related
• The WD statistic is more useful for differentiating between simulated learner types when games are played longer — related
• Segmentation boundary choices differentially affect statistics computed on epistemic-game process data — related
• In the simulation, task complexity and task difficulty explain the majority of variance in the individual-learner WD statistic, with some effect of task specificity — related
• In simulated epistemic games, the weighted density statistic distinguishes simulated learner types with distinct mastery trajectories, with the expert trajectory showing the largest WD values — related
• Across 1.77M answer-pair comparisons, explicit contradiction is rare (0.06%) while compatible differences in detail dominate (57.3% complementary) among non-absent pairs — related
• The system's generated practice trajectories align with learner skill levels at a peak cosine similarity of 0.962 ± 0.014 at the primary difficulty level — related