On inference based on the one-sample sign statistic for long-range dependent data

摘要

This paper considers the problem of statistical inference based on the one-sample sign statistic for strictly stationary random processes which exhibit long-range dependence. Under Gaussian subordination, the limiting distribution of the sign statistic may be non-Gaussian and depends on unknown parameters. We examine how asymptotically valid inference may be carried out using subsampling. The small-sample performance of the method is also investigated by means of Monte Carlo experiments.