Enhancing computational thinking in early childhood education with educational robotics: A meta-analysis

10 studies (15 initially considered)

Early childhood education across Asia, Europe (including Spain) and the Americas.

Severe limitations. The search was restricted to Web of Science and Scopus, to English and Spanish, to OPEN ACCESS only, and to publications since 2019 — a search design that selects for positive open-access results. Only k = 10 survived screening. Heterogeneity is I-squared = 80.7%, and the 95% PREDICTION interval runs from -0.045 to 1.908, meaning a new study could plausibly find nothing. The authors' own bias section is internally contradictory: they report a clean fail-safe N, then state that Begg-Mazumdar and Egger p-values ABOVE 0.05 indicate publication bias, which misreads their own non-significant asymmetry tests.

Pooled SMD 0.9317, 95% CI 0.5969-1.2666, z = 5.4534, p < 0.0001, from ten small open-access studies with SMDs ranging 0.2745 to 1.8272, all positive. Q(9) = 45.93, p < 0.0001, tau-squared = 0.219, I-squared = 80.7%. A pooled effect near 0.93 from ten small studies with 81% heterogeneity and a prediction interval touching zero is a red flag under this archive's benchmarks, not a triumph.

Medium. Mixed designs, small samples, mostly quasi-experimental.

With k = 10 and a prediction interval that includes zero, nothing here is established enough to be replicated or refuted.