Analisi multivariata per osservazioni appaiate con dati mancanti: un caso studio

All parametric approaches require that analysis should be done on complete data sets and so, in presence of missing data, parametric solutions are based either on the so-called deletion principle or imputation methods. But when we delete incomplete vectors we also remove all information they contain, which may be valuable and useful for analysis. And when we replace missing data by suitable functions of actually observed data, that is imputing method, we may introduce biased information which may negatively infuence the analysis. On the other hand, non-parametric solutions in a permutation framework consider data as they are, and units with missing data participate in the permutation mechanism as well as all other units, without deletion or imputing. In this paper we provide a comparison between a parametric solution, represented by ITT principle, and a non parametric one, in a testing problem with multivariate paired observations.