A method of constructing consistent and effective algorithms for robust nonparametric generators of random variables is considered for statistical simulation problems and bootstrap procedures. Semiparametric and semi-nonparametric algorithms of generators have been synthesized for inhomogeneous experimental data. It is shown that standard parametric and nonparametric generators of random variables are inconsistent under conditions of inhomogeneous samples, and their use can significantly and unpredictably distort simulation results and decision-making procedures. In the presence of outliers, the efficiency of the robust semiparametric and semi-nonparametric generators can considerably exceed that of standard generators of random variables, especially in situations with asymmetric outliers