#18, page 258, book 1 Schweser. It shows an AR model, and asks, according to a 5% level of significance, does the model appear to be properly specified? And it calculates the t-statistic with n-2 df. But I thought with an AR model, you don’t use the normal t-statistic for autocorrelation; you use t=(correlations of errors/(1/square root of T)), where T=# of observations and df=T-2. Why are they using a regular t-statistic value when they should use this specific t?
Nopes. To check if any of the autocorrelations of the residuals from the AR model is significant or not we use the normal t-test only.