Theoretical results and empirical evidences indicate that the Ljung–Box test is sensitive to the number of lags ($H$) involved in the test. In time series literature, different values are suggested for $H$. This paper is concerned with the selecting optimal number of lags $H$ in Ljung–Box test such that the actual size of the test does not exceed the test’s level while the power of the test does not fall under specific value. A simulation study is employed to investigate the effect of selecting an improper values of $H$ on the actual size and power of the Ljung–Box test. The results confirm that an optimal value of $H$ depends on the time series’ length as well as the test’s level. The comparison results with currently used approaches in the literature confirm that the commonly used techniques are not suggesting a proper value for $H$.

理论结果和经验证据表明，Ljung-Box检验对滞后次数敏感（$H$）参与测试。在时间序列文献中，建议使用不同的值$H$。本文涉及最佳滞后数的选择$H$在Ljung–Box测试中进行测试，以使测试的实际大小不超过测试的级别，同时测试的功效不低于特定值。通过仿真研究来研究选择不当值的影响。$H$Ljung-Box测试的实际大小和功效。结果证实最佳值为$H$取决于时间序列的长度以及测试的水平。与文献中当前使用的方法的比较结果证实，常用的技术没有表明适当的价值。$H$。