Unbiased Confidence Intervals for Psychological Testing: The Regression-Based True Score Approach With Scale...

Area of Science:

• Psychometrics
• Classical Test Theory
• Statistical Inference

Background:

• Confidence intervals (CIs) are crucial for interpreting individual scores in psychological testing.
• Existing methods, including the traditional and regression approaches, have limitations in accurately reflecting score variability and accounting for regression to the mean.
• Differences in score scaling between observed and true scores complicate CI interpretation.

Purpose of the Study:

• To introduce a novel, scale-corrected regression approach for calculating confidence intervals for individual scores in psychological testing.
• To address the scaling discrepancies inherent in the regression-based true score estimation.
• To provide a more accurate and interpretable method for confidence interval calculation in psychometric practice.

Main Methods:

• The study proposes a scale correction for the regression-based true score estimate.
• Simulations were conducted to evaluate the performance of the new approach against traditional and uncorrected regression methods.
• The corrected regression approach aims to preserve the original score scaling.

Main Results:

• The proposed scale-corrected regression approach demonstrated superior performance in simulations.
• This method effectively addresses the scaling issue, providing CIs on the same scale as observed scores.
• The new approach outperforms both the traditional and uncorrected regression methods in accuracy and interpretability.

Conclusions:

• The scale-corrected regression approach is recommended for calculating confidence intervals in psychological testing practice.
• This method offers a more accurate and practical solution for interpreting individual scores by accounting for measurement error and regression to the mean.
• The findings suggest an advancement in psychometric methodology for score interpretation.