Interval estimation of higher order quantiles. Analysis of accuracy of selected procedures

In the paper selected nonparametric and semiparametric estimation methods of higher orders quantiles are considered. The construction of nonparametric confidence intervals is based on order statistics of appropriate ranks from random samples or from generated bootstrap samples. Semiparametric bootstrap methods are characterized by double bootstrap simulations. The values of bootstrap sample below the prearranged threshold are generated by the empirical distribution and the values above this threshold are generated by the distribution based on the asymptotic properties of the tail of the random variable distribution. The results of the study allow one to draw conclusions about the effectiveness of the considered procedures and to compare these methods.