EViews.com

VAR Residual Serial Correlation LM Tests
Null Hypothesis: no serial correlation at lag order h
Date: 04/26/12 Time: 15:57
Sample: 1961M01 2010M12
Included observations: 594

Lags LM-Stat Prob

1 33.80490 0.1121
2 31.13447 0.1846
3 53.37381 0.0008
4 51.58304 0.0014
5 33.08849 0.1289
6 33.04551 0.1299
7 41.86553 0.0186

Probs from chi-square with 25 df.

Say, if VAR estimated is VAR(6), does this means my VAR model has no autocorrelation problem at all with p=0.1299 or this means my VAR model has autocorrelation problem at lag 3 and 4 but not the other lags?

Help very much appreciated! Also, is the test dependent on normally distributed errors?

Hello, I am not quite sure if there is a response to that but I am wondering the same thing?

If the more than half of the p values exceed 0.05 or 0.1... Then can consider the result is no autocorrelation

I am wondering the same thing.

To the answer "more than half of p values >0,05, then no autocorrelation" ===> I did not meet any such info in literature. Please give your source.

Also, Eviews output sometimes shows VAR Autocorrelation that, p values of some of the lags are <0,05 where some others >0,05.
So, Assume for example VAR(2) is estimated, and VAR Autocorrelation test is done for lags up to 2. The output is:
Lag p
----- -----
1 0,01
2 0,45

Then what will be our decision?
a. we are working with VAR(2), and hence p=0,45>0,05. ==> No autocorrelation for VAR residuals
OR
b. There is a p=0,01<0,05, and hence VAR(2) is problematic in terms of residual autocorrlations?

Any help will be appreciated much.

I am wondering the same thing!!! can any expert send a reply please ?

I have similar problem, also. So, it would be really nice if someone can help us here.

Thanks in advance.