Missing Value Imputation With Adversarial Random Forests-MissARF.

Area of Science:

• Biostatistics
• Machine Learning
• Data Science

Background:

• Missing data is a prevalent issue in biostatistical analyses.
• Imputation methods are standard techniques for addressing missing values.
• Existing methods may vary in efficiency and imputation quality.

Purpose of the Study:

• To propose a novel, fast, and user-friendly imputation method named MissARF.
• To leverage generative machine learning, specifically adversarial random forests (ARF), for imputation.
• To provide both single and multiple imputation capabilities.

Main Methods:

• MissARF utilizes adversarial random forests (ARF) for density estimation and data synthesis.
• Imputation involves conditioning on observed values and sampling from the ARF-estimated conditional distribution.
• The method is designed for both single and multiple imputation scenarios.

Main Results:

• MissARF demonstrates imputation quality comparable to state-of-the-art methods.
• The method achieves a fast runtime, enhancing computational efficiency.
• MissARF provides multiple imputation without incurring additional computational costs.

Conclusions:

• MissARF is an effective and efficient imputation technique for biostatistical analysis.
• The method offers a competitive alternative to existing imputation strategies.
• Its generative machine learning foundation ensures robust data synthesis for missing values.