Item-Level Meta-Analysis of Response Time and Discrimination in 40 Datasets

This dataset comprises 1.84 million item responses aggregated from 40 empirical psychometric datasets. It was created by Josh Gilbert for a meta-analysis investigating conditional dependencies between response time and item discrimination. The analysis found a pooled negative coefficient of -0.27% for the relationship between residual response time and item discrimination.

Use Cases

• Analyze the relationship between residual response time and item discrimination using the pooled coefficient of -0.27% per 1% RT difference.
• Investigate the curvilinear relationship between residual response time and item discrimination, which is maximized just below average RT.
• Examine heterogeneity in the response time-item discrimination relationship across 40 aggregated empirical datasets.
• Model the simultaneous relationships between response time and item discrimination at the person, item, and person-by-item residual levels.

Strengths

• Aggregates 1.84 million item responses for a large-scale analysis.
• Derived from 40 distinct empirical datasets, providing a broad base for meta-analysis.
• Includes specific statistical findings, such as a pooled coefficient of -0.27% with a standard error of 0.04.

Limitations

• The underlying structure and columns of the 40 source datasets are not described.
• High heterogeneity was reported in the analysis, indicating variability across the aggregated studies.
• The dataset is a meta-analytic aggregation, not raw response-level data.

Provenance

Source

Josh Gilbert Dataverse

Collection Method

Item-level meta-analysis aggregating 40 empirical psychometric datasets.

Freshness

Last updated on 2026-02-26.