Principled Missing Data Treatments.

Area of Science:

• Prevention Science
• Biostatistics
• Health Research Methodology

Background:

• Many current missing data practices in prevention research are suboptimal, including listwise deletion and pairwise deletion.
• There is a need to improve the handling of missing data to ensure robust research findings.

Purpose of the Study:

• To promote better practices in handling missing data for intervention and prevention researchers.
• To review current missing data methodology and reporting in prevention research.
• To discuss limitations of antiquated methods and introduce principled approaches.

Main Methods:

• Review of current missing data methodology and reporting in prevention research.
• Discussion of antiquated, ad hoc missing data treatments and their limitations.
• Detailed explanation of two modern, principled missing data treatments: multiple imputation and full information maximum likelihood.

Main Results:

• Antiquated missing data treatments (e.g., listwise deletion) are often ill-advised and limit statistical inference.
• Modern methods like multiple imputation and full information maximum likelihood offer principled approaches to handle missing data effectively.
• These principled methods improve causal and statistical inference in prevention sciences.

Conclusions:

• Prevention researchers should adopt principled missing data treatments such as multiple imputation and full information maximum likelihood.
• Adopting advanced missing data handling techniques is crucial for enhancing the validity of research findings in prevention science.
• Recommendations are grounded in missing data theory and statistical principles for biosocial and prevention research.