Seppo Laaksonen and Pasi Piela
Statistics Finland
FIN-00022 Tilastokeskus
Finland
Email: firstname.secondname@stat.fi
The purpose of the paper is to give an overview of new and traditional methods for imputation, and specify some of these methods with empirical exercises. It is nice that we may approach both to these so-called new techniques and traditional techniques in the same way as described in the attached figure. The difference is derived mainly from a model used prior to imputation. As far as newer techniques are concerned, the imputation model is more exploratory and non-parametric whereas in case of traditional techniques it is parametric and linear. Another difference is concerned the level of automation, traditional techniques being less automated.
Self-organizing maps (SOM) is an iterative method for classification and can thus also be used in finding the imputation classes. Imputations are made within clusters, located by corresponding neurons, in several ways that can be based on both traditional and neural methods as MLP models. Naturally, there are several modifications developed for the SOM, Tree-structured SOM as one of them.
