Principal component analysis with interval imputed missing values

In this paper some statistical properties of Interval Imputation are derived in the context of Principal Component Analysis. Interval Imputation is a recent proposal for the treatment of missing values, consisting of replacing blanks with intervals and then analyzing the resulting data matrix using Symbolic Data Analysis techniques. The most noticeable virtue of this method is that it does not require a single-valued imputation, so it allows us to take into account that incomplete observations are affected by a degree of uncertainty. Illustrative examples and simulation studies are carried out in order to illustrate the functioning of the technique.