Sample partial autocorrelations and portmanteau tests for randomness
This article proposes a new portmanteau test based on sample partial autocorrelations. The test statistic is asymptotically χ2 under the null hypothesis of randomness. Simulation results indicate that the proposed test, which utilizes Anderson's mean and variance formulae of sample partial autocorrelations, outperforms the Ljung-Box test in terms of controlling test size and minimizing dispersion bias.