How Can a Loss of Information in Mixed Attribute Datasets be Prevented?: On the Imputation of Missing Values in Mixed Attribute Datasets Using Higher Order Kernel Functions