In the Schweser example “Testing an AR model for proper specification”, the last sentances say “Additional lags are included in the model and the correlatinos of the residuals (error terms) are checked again. This procedure will be followed until all autocorrelations are insignificant.” Can someone tell me what this means? I would think as one continues to add lags, the number of significant autocorrelations detected can only increase, so I’m confused.
Thanks for the help!
-Patrick
This means that if your running an AR(1) model and the t-test of the residual lag autocorrelations is significant(reject the null hypothesis) then re-specify your model to AR(2) and test again n leep increasing the lag until the t-test the autocorrelations are insignificant.
To simply put it in other models we use DW test for serial correlation and if there is serial correlation we use the hansen methood to correct it.
The same way in AR models we use t-test of the residual lag correlation and re-specify the model by increasing the lags to correct it
read example 6 in the cirriculum to better understand this concept
I hope it helps
Thanks, Hafsa! I was thinking than an additional lag woud be an additional step ahead (i.e. adding a 13th row to the table shown in the example). Now I see adding an additional lag means increasing p: as you say, re-specifying from AR§ = AR(1) to AR§ = AR(2).