Traditional unit weighting (UW) remains ubiquitous in psychological assessment due to its simplicity, yet it assumes equal item contribution and struggles with person-item response inconsistencies, commonly known as the slipping effect. This study introduces the Generalized Conditional Reliability Weighting (G-CRW) algorithm, a parsimonious scoring method for polytomous scales that conditionally incorporates item reliability into observed scores based on a person-item congruence threshold. To evaluate its psychometric performance relative to UW, a comprehensive Monte Carlo simulation (1134 conditions, 1000 replications) and an empirical application (N = 349) using three established scales (Doomscrolling, DASS-21, AAQ-II) were conducted. Simulation results demonstrated that G-CRW yields superior explained variance ratios (EVR) and internal consistency coefficients compared to UW, particularly under normal distributions and high average factor loadings (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\overline{\uplambda }\ge 0.80$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mover> <mi mathvariant="normal">λ</mi> <mo>¯</mo> </mover> <mo>≥</mo> <mn>0.80</mn> </mrow> </math></EquationSource> </InlineEquation>). Confirmatory factor analysis (CFA) fit indices (CFI, TLI, SRMR) favored G-CRW under weaker loading conditions (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$\overline{\uplambda }=0.40$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mover> <mi mathvariant="normal">λ</mi> <mo>¯</mo> </mover> <mo>=</mo> <mn>0.40</mn> </mrow> </math></EquationSource> </InlineEquation>), while method performance converged under highly skewed distributions due to algorithmic functional inertia. Empirical analyses showed that G-CRW scores were highly correlated with UW scores (r > .98) and preserved very similar patterns of associations with external variables, while producing selective score changes for only a subset of respondents. G-CRW provides applied researchers with a computationally efficient, open-source tool for improving psychometric indices without the stringent assumptions of full latent-variable modeling. To ensure immediate applicability and reproducibility, the proposed algorithm is implemented in the open-source WeightMyItems (available at <a href="https://cran.r-project.org/package=WeightMyItems">https://cran.r-project.org/package=WeightMyItems</a>) R package and the user-friendly FAfA Shiny web application (available at <a href="https://cran.r-project.org/package=FAfA">https://cran.r-project.org/package=FAfA</a>).

Mitigating the slipping effect in polytomous scales: The Generalized Conditional Reliability Weighting (G-CRW) Algorithm and the WeightMyItems R package

期刊论文
发表日期：

2026年7月8日

查看全文

Abdullah Faruk Kılıç

摘要

Traditional unit weighting (UW) remains ubiquitous in psychological assessment due to its simplicity, yet it assumes equal item contribution and struggles with person-item response inconsistencies, commonly known as the slipping effect. This study introduces the Generalized Conditional Reliability Weighting (G-CRW) algorithm, a parsimonious scoring method for polytomous scales that conditionally incorporates item reliability into observed scores based on a person-item congruence threshold. To evaluate its psychometric performance relative to UW, a comprehensive Monte Carlo simulation (1134 conditions, 1000 replications) and an empirical application (N = 349) using three established scales (Doomscrolling, DASS-21, AAQ-II) were conducted. Simulation results demonstrated that G-CRW yields superior explained variance ratios (EVR) and internal consistency coefficients compared to UW, particularly under normal distributions and high average factor loadings ( $\overline{\uplambda }\ge 0.80$ $\bar{λ} \geq 0.80$ ). Confirmatory factor analysis (CFA) fit indices (CFI, TLI, SRMR) favored G-CRW under weaker loading conditions ( $\overline{\uplambda }=0.40$ $\bar{λ} = 0.40$ ), while method performance converged under highly skewed distributions due to algorithmic functional inertia. Empirical analyses showed that G-CRW scores were highly correlated with UW scores (r > .98) and preserved very similar patterns of associations with external variables, while producing selective score changes for only a subset of respondents. G-CRW provides applied researchers with a computationally efficient, open-source tool for improving psychometric indices without the stringent assumptions of full latent-variable modeling. To ensure immediate applicability and reproducibility, the proposed algorithm is implemented in the open-source WeightMyItems (available at https://cran.r-project.org/package=WeightMyItems) R package and the user-friendly FAfA Shiny web application (available at https://cran.r-project.org/package=FAfA).