MCAR item and response missingness in linear tests shows no differential effect on model fit across Designs 1 to 4¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2012 · not yet checked against its sources · not reviewed by a person · overdue for a recheck since 2015: tool and model accuracy changes quickly

Subclaims¶

q2 i? Item and response missing data characterized as MCAR for linear tests have no differential effects on model fit across Designs 1 to 4, so linear-test constructs recover reasonably well at the item level. → Shudong Wang 2012

Evidence¶

Shudong Wang 2012¶

Shudong Wang, Hong Jiao. (2012). Examine Construct Validity of Computerized Adaptive Test in K-12 Assessments. Paper presented at the annual meeting of the National Council on Measurement in Education. https://www.nwea.org/research/publication/examine-construct-validity-of-computerized-adaptive-test-in-k-12-assessments/

q2 · i? · causal · r2

Simulation results (Tables 7 and 8) comparing model fit across Designs 1-4, which vary MCAR item and response missingness; the article reports "Both unidimensional models (Rasch and 2PL) fit data well" while the bifactor model recovered data poorly.

"First, at the item level, results show that both item and response missing data characterized as MCAR for linear tests have no differential effects on models fit across Designs 1 to 4. Both unidimensional models (Rasch and 2PL) fit data well but the multidimensional model (bifactor model) recoveres data poorly."

Discussion¶

Related Claims¶

• CAT data cannot be fit into factor models at the item level, making item-level construct validity investigation meaningless for CAT — related
• Existing item fit statistics are derived for linear tests under assumptions of no missing responses and normal ability distributions, leaving adaptive testing contexts understudied — related
• Item parceling at the testlet level substantially improves model fit, allowing partial (Rasch, 2PL) or full (bifactor) recovery of CAT constructs — 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
• Missing data in the analysis samples occurred largely at random — related
• The trifactor model's performance across sample sizes and model specifications was examined via simulation studies comparing model fit, bias, and convergence rates — related