Statistical primer: how to deal with missing data in scientific research?

Area of Science:

• Biostatistics
• Data Science
• Scientific Research Methodology

Background:

• Missing data frequently challenge research integrity and statistical inference.
• Inappropriate handling of missing data can lead to biased results and compromised conclusions.
• Understanding and effectively managing missing data is essential for robust scientific findings.

Purpose of the Study:

• To introduce fundamental concepts of missing data to a non-statistical audience.
• To compare popular methods for handling missing data in research.
• To provide guidelines for managing and reporting missing data in scientific studies.

Main Methods:

• Review and comparison of common missing data handling techniques.
• Explanation of complete case analysis and single imputation.
• Detailed discussion and application of multiple imputation for data missing at random.
• Illustration using Cox regression analysis.

Main Results:

• Complete case analysis and single imputation are widely used but often yield invalid inferences.
• Multiple imputation is a powerful and flexible method for data missing at random.
• Careful implementation of multiple imputation is necessary for reliable results.

Conclusions:

• Effective handling of missing data is vital for research validity.
• Multiple imputation is recommended as a robust approach for missing data, surpassing simpler methods.
• Adherence to guidelines for managing and reporting missing data ensures research transparency and integrity.