Comparative methods for handling missing data in large databases

Missing data are an important analytic topic in research with large databases. The commonly used missing indicator variable method introduces severe bias and should be used with caution. We present empiric evidence to guide method selection for handling missing data.