Mecor: An R package for measurement error correction in linear regression models with a continuous outcome

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Measurement error in covariates or outcomes is prevalent in regression models.
• Ignoring this error can lead to significant bias in estimated associations.
• Existing measurement error correction methods are underutilized.

Purpose of the Study:

• To develop an R package, mecor, to facilitate the application of measurement error correction methods.
• To implement regression calibration and maximum likelihood methods for covariate error.
• To implement methods of moments for outcome error.

Main Methods:

• The mecor R package implements various measurement error correction techniques.
• Methods include regression calibration and maximum likelihood for continuous covariates.
• Methods of moments are included for continuous outcomes.
• Parameter estimation utilizes data from validation, replicates, calibration, or external studies.
• Variance estimation is provided via closed-form solutions and bootstrapping.

Main Results:

• The mecor package provides accessible tools for correcting measurement error in regression models.
• It supports correction for errors in both continuous covariates and outcomes.
• The package offers robust variance estimation for corrected estimators.

Conclusions:

• The mecor R package aims to increase the uptake of measurement error correction methods in statistical analyses.
• It provides a practical solution for researchers dealing with measurement error in regression.
• Improved handling of measurement error leads to less biased covariate-outcome association estimates.