How to interpret the results of Breusch Godfrey test

[bbmra]

Hi,

I am calling Breusch Godfrey api to test autocorrelation and got results like the follows

Breush Godfrey: lm= 7.85128552435 , lmp= 0.0971802297362 , fv= 1.9536984474 , fp= 0.10221325756

with 95% confidence level should I take it as no autocorrelation or not.

Many thanks for any comments in advance.

[josef...@gmail.com]

With pvalues of around 0.1, we would not reject the null hypothesis of no serial correlation. There might be some small indication of autocorrelation at a type 1 error of 0.1.

How you interpret or use this depends on your sample, e.g. sample size, and on what you use it for.

If this are OLS residual and the interest is in inference on the parameters, I would check that the cov_type='HAC' autocorrelation robust standard errors don't differ much from the OLS standard errors, and that no substantial conclusion depends on it.

If this are the residual of an ARIMA model, then I would check larger ARIMA models if the sample size is large enough, i.e. enough degrees of freedom, but I would ignore it (not rejecting the model specification) with a small sample size.

Josef