Nonparametric Simultaneous Confidence Intervals For Mean Vector

We propose a new method of constructing simultaneous confidence intervals with marginal equal tail probability for the mean vector. By the bootstrap estimation of mean vector, we use the percentile interpolation to construct simultaneous confidence intervals. We compare our new method with the Bonferroni method, Efron’s method, Normal Exact method and Bootstrap simultaneous confidence intervals. Simulation results show that the proposed simultaneous confidence intervals are comparable to the other bootstrap methods in terms of coverage probability. In addition, it has the property of marginal equal tail probability and is simple and feasible.