A true score imputation method to account for psychometric measurement error

Area of Science:

• Psychometrics
• Statistical Modeling
• Data Analysis

Background:

• Self-report questionnaire scores are frequently used in statistical models without accounting for measurement error.
• This omission leads to biased estimates in statistical analyses.
• Existing measurement error corrections have limitations in applicability or complexity.

Purpose of the Study:

• To introduce a flexible multiple imputation-based approach, true score imputation, to address limitations of existing measurement error corrections.
• To provide a method that accommodates a broad class of statistical models.
• To offer a unified framework for handling both missing data and measurement error.

Main Methods:

• True score imputation augments datasets with plausible true scores.
• The augmented datasets are analyzed using standard multiple imputation methodology.
• A simulation study and a real-world data example were used for evaluation.
• An R package is available for implementing the method within the 'mice' library.

Main Results:

• True score imputation significantly reduces bias compared to analyses that ignore measurement error.
• The method provides accurate point estimates and confidence intervals.
• The approach is compatible with existing multiple imputation techniques for missing data.

Conclusions:

• True score imputation offers a flexible and effective solution for accounting for measurement error in statistical analyses.
• The method enhances the accuracy of statistical estimates derived from questionnaire data.
• This approach integrates measurement error correction with missing data handling for comprehensive data analysis.