Comment on Oberman & Vink: Should we fix or simulate the complete data in simulation studies evaluating missing data...

Area of Science:

• Statistics
• Data Science
• Computational Statistics

Background:

• Simulation studies are crucial for evaluating statistical methods, particularly for handling missing data.
• A common approach involves generating partially observed data from complete datasets.
• Simulating missingness indicators after fixing complete data is a frequently considered, yet often flawed, technique.

Discussion:

• The method of fixing complete data and simulating missingness indicators can lead to biased evaluations of missing data handling techniques.
• This approach may not accurately represent the mechanisms that cause data to be missing in real-world scenarios.
• Researchers must carefully consider the appropriateness of this simulation strategy to avoid misleading conclusions.

Key Insights:

• Generating partially observed data by simulating missingness indicators is only rarely appropriate for missing data simulation studies.
• This simulation technique can produce superficially attractive but ultimately misleading results.
• The validity of simulation studies hinges on the accurate representation of missing data mechanisms.

Outlook:

• Future research should focus on developing and validating more appropriate simulation strategies for missing data.
• Emphasize the importance of understanding and correctly modeling missing data mechanisms in simulation studies.
• Promote rigorous methodological standards for evaluating statistical techniques in the presence of missing data.