Autor: Rodrigo Alfaro; Marcelo Fuenzalida.

Cuadernos de Economía, Vol. 46, N° 134, pp. 273-288, 2009.

In the survey analysis, the missing data problem can be managed by using Multiple Imputation (MI) methods. In this paper we show the empirical application of MI methods to the financial variables included in Chile’s Social Protection Survey 2004. Based on a brief review of MI methods we conclude that Multivariate Normal one is more appropiate for our case. In addition, we consider two empirical adjustments: (1) use of the variables in their logistic versions, and (2) implementation of the method by groups of individuals. Our results show that both adjustments improves the performance of the MI method.

Clasificación JEL: C11, C15

Palabras Clave: Información Faltante (Missing Data), Imputación Múltiple, Algoritmo EM/DA.