The 2005 Plenary Meeting on 'Missing Data and Measurement Error'

摘要

The treatment of missing data in statistical analysis is one of the most disliked topics in statistics. Although a lot of methodological progress has been reached since Rubin's definition of "missing completely at random" (MCAR), "missing at random" (MAR) and "not missing at random" (NMAR) in 1976, ignoring incomplete cases still seems to be the most popular strategy. Milestones of the methodological progress were collected by the "Panel on Incomplete Data" (Madow, et al. 1983) from the view of Official Statistics. Here emphasis was given to weighting as a method to compensate for missing data. A more recent milestone was the volume on survey nonresponse edited by Groves, et al. (2002).