Distribution-free multiple imputation from a deterministic simple imputation method

The contemporary statistical literature suggests the application of multiple imputation (MI) to solve the problem of missing values in experimental data and data set with public or private character. The theory of missing values and the associated statistical software provide several alternatives of MI, for instance, parametric regression, the propensity score method and the Markov Chain Monte Carlo (MCMC) method, however, these methodologies require that certain assumptions are met, both at the structural (random mechanism of missing values) as in the distributional level (multivariate normality). When these assumptions are not satisfied, the methods can provide unexpected results, because of that, arises the necessity of alternatives non-parametric without structural requirements.