Multiple Uses of (Multiple) Imputation at the National Center for Health Statistics
Nathaniel Schenker, Senior Scientist for Research and Methodology, National Center for Health Statistics, Hyattsville, MD
Monday, February 25, 2008 11:00 AM - 1:00 PM
To illustrate several types of problems in which imputation for missing data can be used, this talk will describe several recent or potential applications of imputation at the National Center for
Health Statistics, Centers for Disease Control and Prevention. Two of the applications are to traditional missing-data problems, one involving missing income data in the National Health Interview
Survey and the other involving missing data on body scans in the National Health and Nutrition Examination Survey. Two other applications have the goal of imputing values for an item of interest not
included in a survey by using responses to a related item along with models based on a smaller "validation" survey. One such application involves bridging the transition from single-race reporting to
multiple-race reporting in the U.S. census, and the other seeks to improve on analyses of self-reported data from the National Health Interview Survey. The final application uses mixture models to
identify questionable birth weight/gestational age pairs in U.S. natality data, for which gestational ages can be measured with error, and seeks to impute more plausible gestational ages. In all of
the applications, multiple imputation or related techniques are used to reflect the extra variability due to missing data. The talk will provide a brief introduction to imputation and multiple
imputation and will highlight goals, techniques, and results of the applications. Dr. Schenker has served at the US Census Bureau, the UCLA School of Public Health, and the National Center for Health
Statistics. He is an expert on multiple imputation and is an outstanding public speaker.
