I want to test my data for no serial correlation, and of course I use the Durbin Watson statistic. This statistic is given by my econometric software (TSP) when estimating the data. And I compare this statistic with the so-called "du" and "dl" values in the literature.
As far as I know, if the statistic is above "du", I can confidently reject the autocorrelation. If it is under, I cannot conclude and if it is under dl I should conclude there is positive autocorrelation.
OK, no problem so far.
But I cannot understand the "du" values for different confident intervals. An example is best to illustrate my problem.
If I wish an 1% confidence value, the literature displays du=1.43 (32 observations, 3 variables + intercept.
But if I wish a 5% confidence, the literature displays du=1.65
Suppose now I have a statistic equal to 1.5. This is above 1.43 thus there is no autocorrelation with an 1% confidence value
But 1.5 is under 1.65 thus I cannot have the same conclusion although I have made the confidence looser.
Isn't that queer ?
Is there an explanation or am I wrong with my confidence hypothesis ?
For econometric discussions not necessarily related to EViews.
1 post • Page 1 of 1
Who is online
Users browsing this forum: No registered users and 4 guests