Abstract:
|
Standard applications of multiple imputation (MI) techniques are based on the assumption that the data are Missing at Random (MAR). However, in many situations it seems very realistic that the missing values follow a Missing Not at Random (MNAR) mechanism. In this case, usual implementations of MI are not sufficient either and may lead to biased estimates. We will present an approach to multiply impute non-ignorable binary missing data in the framework of Fully Conditional Specification (FCS). The suspected MNAR mechanism will be considered and modeled during the imputation process by applying a censored bivariate probit model as imputation model. For allowing the consideration of a present multilevel structure in the data during the imputation process, the model is expanded by a random intercept term. In order to assess the performance of this imputation technique, different simulation studies were conducted. The method performed well in all considered situations in terms of coverage and bias and outperforms alternative methods. Finally, in order to evaluate its applicability the approach is employed on empirical data of the National Educational Panel Study (NEPS).
|