Multiple imputation in quantile regression.

Area of Science:

• Statistics
• Biostatistics

Background:

• Missing covariates in statistical models can lead to biased and inefficient parameter estimates.
• Quantile regression is a powerful tool for analyzing the conditional distribution of response variables.
• Robust estimation methods are crucial for handling real-world data complexities.

Purpose of the Study:

• To develop and evaluate a multiple imputation estimator for quantile regression with missing covariates.
• To enhance parameter estimation efficiency and consistency in the presence of missing data.
• To introduce a shrinkage estimator to mitigate potential bias from model misspecification.

Main Methods:

• Proposed a multiple imputation technique tailored for quantile regression models.
• Developed a shrinkage estimator to address potential model misspecification.
• Conducted a simulation study to assess the finite sample performance of the proposed estimator.
• Applied the methodology to real-world dietary intake data from the Eating at American's Table Study.

Main Results:

• The multiple imputation estimator demonstrated increased efficiency by utilizing the entire dataset.
• Coefficient estimators were found to be root-n consistent and asymptotically normal.
• The shrinkage estimator effectively adjusted for possible bias, improving robustness.
• The methodology was successfully applied to investigate associations in dietary intake data.

Conclusions:

• The proposed multiple imputation and shrinkage estimators provide an efficient and robust approach for quantile regression with missing covariates.
• The methods are suitable for handling complex datasets and can improve the reliability of statistical inference.
• The application to dietary data highlights the practical utility of the developed methodology in health and nutrition research.