Optimal Lag for Breusch–Godfrey Test

[Nick Mueller]

Dear all,

I have a question about the optimal number of lags for the Breusch-Godfrey Test. I have daily data of US Bond yields for two and ten years. Due to the fact that I am suspecting serial correlation, I would like to test with the aforementioned test, however, I do not know which lag length to use.

Does anybody have a rule of thumb or method on how to determine the optimal lag length?

Any hint is appreciated!

Best,
Nick

[Justin Niakamal]

You can specify the lags and see the test results at different lag orders using the lags() option.

webuse klein, clear
tsset yr
regress consump wagegovt L.consump L2.consump
estat bgodfrey, small lags(1/10)

Breusch-Godfrey LM test for autocorrelation
---------------------------------------------------------------------------
    lags(p)  |          F                  df                 Prob > F
-------------+-------------------------------------------------------------
       1     |        0.107           (  1,   15 )              0.7484
       2     |        0.358           (  2,   14 )              0.7056
       3     |        0.286           (  3,   13 )              0.8346
       4     |        2.216           (  4,   12 )              0.1285
       5     |        1.889           (  5,   11 )              0.1762
       6     |        1.622           (  6,   10 )              0.2380
       7     |        1.421           (  7,    9 )              0.3049
       8     |        1.244           (  8,    8 )              0.3824
       9     |        2.110           (  9,    7 )              0.1687
      10     |        1.899           ( 10,    6 )              0.2233
--------------------------------------------------------------------

Hope this helps.