Sensitivity analyses showed most sequential CVT paths were spurious due to confounders; only interest→enjoyment was robust¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r2 · q2 · i0 negligible
Currency · newest study 2023 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Nearly all paths identified in the reverse causality models were no longer significant when coevals and covariates were controlled for (ps ≥ .11), whereas the interest→enjoyment link was highly robust to unobserved confounders (ρ(U) = .31). → Chevalère 2023 q3 i0 The only Forward path unbiased by observed confounders, performance T6 → final performance → learning gains (β = .08), was sensitive to unobserved confounding, abolished at ρ(U) = .18. → Chevalère 2023 (2)

Evidence¶

Chevalère 2023¶

Chevalère, J., Yun, H. S., Henke, A., Pinkwart, N., Hafner, V., & Lazarides, R. (2023). A sequence of learning processes in an intelligent tutoring system from topic-related appraisals to learning gains. Learning and Instruction, 87, 101799. https://doi.org/10.1016/j.learninstruc.2023.101799

q2 · i? · associational · r2

Mediation-based sensitivity analyses (Imai et al. framework, 5,000 bootstrapped samples) broke each sequence into simple mediation models conditioning on observed coevals; the article states "nearly all paths identified in the reverse causality models (in bold) contained spurious associations".

"the analyses revealed that these paths were no longer significant when coevals and other covariates were controlled for, ps ≥ .11"

Chevalère 2023 (2)¶

Chevalère, J., Yun, H. S., Henke, A., Pinkwart, N., Hafner, V., & Lazarides, R. (2023). A sequence of learning processes in an intelligent tutoring system from topic-related appraisals to learning gains. Learning and Instruction, 87, 101799. https://doi.org/10.1016/j.learninstruc.2023.101799

q2 · i0 · associational · r2

For the sole unbiased Forward path, the mediation model "proved sensitive to the presence of an unobserved confounder U", with the indirect effect abolished at ρ(U) = .18, below the article's robust threshold of .20.

"the unique variance of performance 6 predicted learning gains through the final performance score, β = . 08, SE = .04, t = 2.03, p = .030 [.001; .18]"

Discussion¶

Related Claims¶

• The ESDM's unidirectional trait-ACH-to-trait-ASC path is not empirically identified, and the confounder-based RI-LTVC alternative shows poor fit with no support for a time-varying confounder — related
• Reanalyses of simulated and PALMA data provide converging support for the extended reciprocal effects model, including an effect of academic self-concept on achievement — related