Relatore
Riccardo D’Alberto
Abstract
Non-parametric micro Statistical Matching imputation techniques (the so-called “hot deck” methods) are commonly applied to integrate data and overcome commonly occurring issues such as elementary data shortage and missing data.
These methods received an increasing attention in the last decade due to their flexible use and the computational advantages w.r.t. the parametric ones, but also because they allow researchers to avoid model misspecification bias. Further developments and theoretical advances of hot deck methods concern new combinations of the existing techniques and not-default distance functions but also and almost the assessment of the imputation goodness.
Approaching some methodological challenges, a newly-proposed validation strategy for the assessment of the imputation goodness is showed and discussed, taking into account new combinations of techniques and distance functions.
Organizzazione
Alessandra Luati, Silvia Cagnone