2

It appears the Ljung-Box test is not applicable on residuals from autoregressive models, including ARMA; see "Testing for autocorrelation: Ljung-Box versus Breusch-Godfrey". But is it applicable on residuals from MA(q) models?

Richard Hardy
  • 67,272
  • There is a related question on testing the same thing with the Breusch-Godfrey test. – Richard Hardy Dec 14 '20 at 12:51
  • I too have a similar additional doubt: If the Ljung-Box test and Breusch-Godfrey test, both can be applicable for MA models, then is there a significant difference between the 2 in that scenario? Or which is preferable in that case?

    My assumption is that based on the answer from Alecos in above linked question, with MA model, all regressors can be assumed to be uncorrelated to the error terms under the null, so strict exogeneity for Ljung-Box test would hold true. But I am not sure, which holds better, when compared to B-G test.

     – Chintan Rajvir Mar 09 '21 at 05:00
  • 1
    @ChintanRajvir, consider posting this as a new question. – Richard Hardy Mar 09 '21 at 06:36
  • I have added a similar question here: https://stats.stackexchange.com/questions/513136/using-ljung-box-test-vs-breusch-godfrey-test-in-ma-arima-models. – Chintan Rajvir Mar 10 '21 at 04:54

0 Answers0