Latent singing ability varies widely across learners, with sequence-level theta estimates spanning −3.79 to 3.36 logits¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2031: baselines and conditions change

Subclaims¶

q2 i? The IRT fit recovered a wide ability spread (SD 1.06), with about 12.4% of estimates more than one SD below and 13.7% more than one SD above the mean. → Wei 2026

Evidence¶

Wei 2026¶

Wei, Wang and Dong. (2026). Cognitive and skill acquisition trajectories in school-based music education: evidence from Chinese classrooms. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1905847

q2 · i? · associational · r2

IRT parameter estimation (Figure 5) at the sequence level for 2,432 sequences, with a median standard error of 0.157 from the log-likelihood curvature. Pitch difficulties had a much narrower spread (mean −0.74, SD 0.38).

"The IRT fit recovered a wide spread of latent abilities, ranging from -3.79 to 3.36 logits with a standard deviation of 1.06"

Discussion¶

Related Claims¶

• A one-parameter IRT model outperforms global and per-pitch baselines on AUC, BCE and Brier score but not on threshold accuracy for held-out singing attempts — related
• Low-register pitches are sung more accurately than high-register pitches in Chinese classrooms, reproducing register compression at scale — related
• A semi-supervised deep learning architecture recovers the ability distribution of anchor students more accurately than direct 2PL-IRT fitting under simulated nonignorable missingness — related