The RATS Software Forum

Hi Tom,

I am trying to test for serial auto correlation in the residuals after estimating the var. I used

@mvqstat(lags=p)
# residuals

(where p=4)

And I get the output as

Multivariate Q(4)= 27.54972
Significance Level as Chi-Squared(100)= 1.00000

I guess that the first line shows the Q statistic with the number of correlations tested (here 4). The second shows the significance level and the degrees of freedom.
So what would this mean then? That there's no serial correlation in the residuals?
Also what if I want to perform the BG LM test for serial correlation on RATS for the VAR ? Could you tell me how to do that?

Many Thanks

RK2509 wrote:Hi Tom,

I am trying to test for serial auto correlation in the residuals after estimating the var. I used

@mvqstat(lags=p)
# residuals

(where p=4)

And I get the output as

Multivariate Q(4)= 27.54972
Significance Level as Chi-Squared(100)= 1.00000

I guess that the first line shows the Q statistic with the number of correlations tested (here 4). The second shows the significance level and the degrees of freedom.
So what would this mean then? That there's no serial correlation in the residuals?

Right. 4 correlations, 5 variables means 100 correlations and cross correlations. The null is no correlation, so yes, you would conclude that there's no (joint) serial correlation.

RK2509 wrote: Also what if I want to perform the BG LM test for serial correlation on RATS for the VAR ? Could you tell me how to do that?

You rerun the VAR adding lags (however many you want to test) of all residuals.

Many Thanks Tom.