Manual Breusch Godfrey test after ARIMA

[Shannon Murray]

Hi all,

I was wondering if somebody could talk me through how to perform a manual breusch godfrey test after estimating an ARIMA model that contains MA and seasonal terms.
I understand how to estimate the model residuals i am just unsure of the auxiliary regression test equation and the arima/sarima commands

For example,

I run the following model commands:

arima y if tin(..), arima(0,1,1), sarima(1,1,0,12) ( i have monthly data)

b/c i have monthly data i assume i want to check for serial-correlation of order 12? but how do i tell stata this?

If anybody could shed some light it would be greatly appreciated!! i can't find any guidance anywhere online


Thanks

[Justin Niakamal]

See help arima postestimation to see what's available. The reason you're not having luck is because it doesn't make a lot of sense to do what you're asking. Including ar terms will generally take care of any serial correlation issues. You could take a look at the correlogram of the residuals doing something like:

arima y, arima(0,1,1), sarima(1,1,0,12) 
predict r, resid
corrgram r, lags(12)

[Shannon Murray]

Thanks for the reply Justin, i have tried the stata manual but it does not have what i need.

I'm doing a project for uni and it requires that we perform this manual BG test so that we can confirm we have removed all serial correlation from residuals i.e. they are white noise.

Problem is i do not now how to construct the auxiliary regression and enter it into stata (taking into consideration the ma and seasonal terms)

[Justin Niakamal]

i have tried the stata manual but it does not have what i need.

In general, there usually is a good reason for something not being readily available. In my view, a BG test on an arima model is a bit odd. I would refer to the Q-statistics from the correlogram of the residuals.