Abstract
This article concerns item nonresponse adjustment for two-stage cluster samples. Specifically, we focus on two types of nonignorable nonresponse: nonresponse depending on covariates and underlying cluster characteristics, and depending on covariates and the missing outcome. In these circumstances, standard weighting and imputation adjustments are liable to be biased. To obtain consistent estimates, we extend the standard random-effects model by modeling these two types of missing data mechanism. We also propose semiparametric approaches based on fitting a spline on the propensity score, to weaken assumptions about the relationship between the outcome and covariates. These new methods are compared with existing approaches by simulation. The National Health and Nutrition Examination Survey data are used to illustrate these approaches.
Original language | English (US) |
---|---|
Pages (from-to) | 1172-1180 |
Number of pages | 9 |
Journal | Biometrics |
Volume | 63 |
Issue number | 4 |
DOIs | |
State | Published - Dec 2007 |
Keywords
- Cluster-specific nonignorable nonresponse
- Item nonresponse
- Outcome-specific nonignorable nonresponse
- Penalized spline of propensity prediction
- Two-stage cluster sample
ASJC Scopus subject areas
- Statistics and Probability
- General Biochemistry, Genetics and Molecular Biology
- General Immunology and Microbiology
- General Agricultural and Biological Sciences
- Applied Mathematics
MD Anderson CCSG core facilities
- Biostatistics Resource Group