The classical bootstrap and similar methods are widely used to solve statistical problems. This paper considers four resampling methods specially tailored for the data consisting of many interval-valued variables. Two of these approaches are generalizations of the classical and smoothed bootstrap for such a particular case. The following two are non-parametric approaches that take into account possible dependencies within intervals for each variable and between the variables themselves. These algorithms are compared using numerical simulations, various error measures, and statistical tests to check their overall quality. It seems that particularly one of the considered resampling methods gives valuable bootstrapped samples.