Lambda-OR: Replication Data for Misclassification-Adjusted Odds Ratios

Ishanu Chattopadhyay released this replication dataset in 2026 to support the Lambda-OR framework, a statistical method for correcting odds ratio bias in large observational studies. The data facilitates the estimation of true effect sizes when outcomes are defined by imperfect proxy labels or predictive models rather than gold-standard cases. It specifically addresses misclassification issues common in electronic health records and large-scale screening pipelines.

Use Cases

• Correcting bias in odds ratio estimates using the Lambda-OR misclassification operator
• Adjusting 2x2 contingency tables for exposure and proxy outcomes in electronic health records
• Estimating variance for misclassification-adjusted log odds ratios using validation cohort uncertainty

Strengths

• Implements ridge-stabilized inversion to ensure non-negative reconstructed counts
• Accounts for uncertainty in both sampling variability and proxy model operating characteristics
• Provides a principled correction for bias in large-scale observational cohorts

Limitations

• Methodological effectiveness is dependent on the availability of sensitivity and specificity estimates from a validation cohort
• Restricted to binary outcome analysis within 2x2 contingency tables
• Lack of detailed column metadata limits immediate interpretability without the accompanying software

Provenance

Source

Ishanu Chattopadhyay Dataverse

Collection Method

Statistical replication data derived from the Lambda-OR framework development

Freshness

Last updated March 2026.