Efficient Standard Error Formulas of Ability Estimators with Dichotomous Item Response Models

Area of Science:

• Psychometrics
• Statistical modeling
• Educational measurement

Background:

• Item response theory (IRT) models are widely used for ability estimation.
• Accurate estimation of standard errors is crucial for reliable ability estimates.
• Existing methods for asymptotic standard errors (ASE) have limitations for certain estimators.

Purpose of the Study:

• To develop a unified framework for computing ASE of ability estimators in dichotomous IRT models.
• To derive new ASE formulas for weighted likelihood and maximum a posteriori estimators.
• To compare the performance of new ASE formulas against traditional ones.

Main Methods:

• Utilized M-estimation theory to derive a general formula for ASE.
• Applied the framework to maximum likelihood, weighted likelihood, maximum a posteriori, and robust estimators.
• Conducted a simulation study under Rasch modeling to evaluate proposed ASE formulas.

Main Results:

• A general ASE formula was derived, encompassing existing methods.
• New ASE formulas were proposed for weighted likelihood and maximum a posteriori estimators.
• The simulation study provided empirical evidence for the proposed formulas' validity.

Conclusions:

• The M-estimation framework offers a flexible approach to calculating ASE for various ability estimators.
• The newly proposed ASE formulas enhance the accuracy of standard error estimation in specific IRT contexts.
• This work contributes to more reliable ability estimation in educational and psychological assessments.