Typical survey data: Estimation and imputation - ARCHIVED
Articles and reports: 12-001-X197600100002Description: A special class of missing data problems is discussed, namely that of typical survey data whereby zeros dominate the multivariate response space. Here, techniques which impute means (whether conditional or unconditional) distort rather than improve the quality of the data. A probabilistic model is described which provides reasonable estimates, but also upholds the integrity of the data base. Results are given from a comparative study of the proposed methodology with other estimation/imputation models.
Issue Number: 1976001Author(s): Pregibon, DarylMain Product:Survey Methodology
