Alternative Variance-Ratio Tests Using Ranks and Signs
提出基于时间序列秩和符号的方差比检验，用于检验序列是否为鞅差序列。蒙特卡洛模拟显示其比传统方差比检验更精确、更有效，应用于五组汇率数据能检测到鞅假设的违反。
This article proposes using variance-ratio tests based on the ranks and signs of a time series to test the null that the series is a martingale difference sequence. Unlike conventional variance-ratio tests, these tests can be exact. In Monte Carlo simulations, I find that they can also be more powerful than conventional variance-ratio tests. I apply the proposed tests to five exchange-rate series and find that they are capable of detecting violations of the martingale hypothesis for all five series, whereas conventional variance-ratio tests yield ambiguous results.