Abstract:
|
Missing data are frequently encountered either by chance or design. A naive analysis with complete cases can lead to biased estimation and inefficiency. Imputation is a process of assigning values to the missing items with the objective of reducing bias and improving the efficiency. In survey, fractional imputation (FI) is attractive in three fold. First, it constructs imputed values with fractional weights which facilitates full-sample estimators. Second, it allows consistency among different users. Third, FI procedure furnishes good estimates of distribution function and the resulting fractionally imputed data meet the goal of multiple use. In more general disciplines, the FI method can be a substitution for a computationally difficult or intractable expectation step in the EM algorithm. We demonstrate the empirical relevance of FI using simulation designs, compared to the multiple imputation method.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.