Durbin-Watson Test

The Durbin-Watson statistic is produced with every time-series regression estimated by least squares. To access the p-value associated with the test, which is computed using the Imhoff procedure, use the accessor $dwpval. An example based on the Phillips curve is:

1 open "@gretldirdatapoephillips_aus. gdt"

2 diff u

3 setinfo inf – d "Australian Inflation Rate" – n "Inflation Rate"

4 setinfo d_u – d "Change in Australian Civilian Unemployment Rate

5 (Seasonally adjusted)" – n "D. Unemployment Rate"

6 ols inf d_u const

7 scalar dw_p = $dwpval

8 print dw

The result, including the last line of the regression output that shows the estimated value of p and the DW statistic, is:

Подпись: rho0.549882 Durbin-Watson 0.887289

dw_p = 2.1981736e-009

image327

The DW statistic is 0.887 and its p-value is well below the 5% threshold, indicating significant autocorrelation. The GUI gives a slightly prettier result. It has to be called from the model window as Tests>Durbin-Watson p-value.

Many interpret a significant DW statistic as evidence of general model misspecification.

Leave a reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>