Gestion Des Donnees Manquantes Dans Les Bases De Donnees En Sciences Sociales : Algorithme Nipals Ou Imputation Multiple?

Abstract

The main objective of this paper is to assess the robustness of imputation methods to fill up the series of secondary data in social sciences. The methodology used, especially that of mean imputation, multiple imputation and NIPALS algorithm, is based on a simulation using observed data. Results show a close similarity between the observed data and the data obtained by multiple imputation, mean imputation and NIPALS algorithm. The results also suggest that multiple imputation provides values substantially similar to observed data.