Variances of Imputation Variances as Determiner of Sufficient Number of Imputations Using National Ambulatory Medical Care Survey as Model

• Activity Number: 75
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Government Statistics Section
• Author(s): Qiyaun Pan*+ and Rong Wei and Iris M. Shimizu and Eric Jamoom
• Companies: CDC/NCHS and NCHS/CDC and CDC/NCHS and CDC/NCHS
• Keywords: Multiple imputation ; sufficient number of imputations ; hot-deck imputations ; National Ambulatory Medical Care Survey (NAMCS)

Abstract

How many imputations are sufficient in multiple imputation? The answer given by different researchers varies from as few as 2 to as many as hundreds. Most of these conclusions were based on mathematic simulations with various assumptions. It is questionable whether any of these recommendations would fit all data situations. In this research, we determined the sufficient number of imputations based on the relationship between the number of imputations and the variance of the between-imputation variances. We used the data from the 2012 Physician Workflow Supplement of the National Ambulatory Medical Care Survey as the model. Hot-deck based imputations were carried out for 5 variables from the survey, with one being the number of clinic staff and the other four being the practice size and its derivatives. Our results suggest that 20 to 40 imputations are needed to produce a statistically reliable estimate of imputation variances. The method described in this paper may have reference value in determining the sufficient number of imputations for other surveys.