I understand that significant large values (>2) of the Durbin Watson statistic point to negative serial correlation.
I thought that the DW test could just tell us whether or not there is statistically significant positive serial correlation, with the decision rules as follows:
- Reject H0 : means positive serial correlation
- Fail to reject H0 : no evidence of positive serial correlation > so it could be either no serial correlation or negative serial correlation (correct?)
But I found the following in the correction of the topic tests (in Reyes): "Specifically, if the DW statistic exceeds 4 – dl, where dl is the lower critical value of the Durbin–Watson test, there is significant negative serial correlation. "
Where does this come from? Can anyone confirm/clarify?