I am actually (comparing) fitting a particular time serial with nonlinear (SETAR) model and linear (AR) model. During Model Diagnostics, the SETAR model gave the the p-value and its graph; while AR model did not. For the AR model, the R programming I used reported that the p-value and its graph could not be estimated and plotted since the psi weights are not small enough for the Ljung-Box tests.I wish to understand the meaning and implication of that note. A material on "psi weights in Ljung-Box tests" could be helpful. Thank you, Thomas Walther for your concern.