ABSTRACT
Missing data is a ubiquitous problem in the analysis of survey data. Section 2.7.3 introduced the concept of unit nonresponse and weighting adjustments to compensate for potential bias due to completely missing data for significant fractions (e.g., 10%, 20%, 30%) of the probability sample that was selected to represent the survey population (see Figure 11.1). The example analyses of the National Comorbidity Survey Replication (NCS-R), 2006 Health and Retirement Survey (HRS), and 2005-2006 National Health and Nutrition Examination Survey (NHANES) data sets presented in Chapters 5 through 10 all employed weights that adjusted for unit nonresponse on the part of original sample.
