2

Isn't this test for the determination of auto-correlation of residuals only necessary when time is some sort of a factor in the observed variables?

As it is I had a data-set that had one dependent variable (sales) and three independent variables, (unemployment rate, population size and advertizing expense). I find it difficult to imagine how the residuals in this situation can be serially-related without a sequence of observations being specified.

My group-mate was unable to explain this to me as I was very skeptical that the Durbin Watson test belonged in our report.

Please help me understand

Siyanda
  • 23
  • 1
  • 3

1 Answers1

1

You can also use this test to detect spatial autocorrelation. A random shock affecting sales in one region may also cause sales in an adjacent region to change because of close economic ties between them. Weather shocks are another example. I once worked on a project where there was a Superbowl effect, where the sales spiked near the home cities of the two teams.

But you have to be careful, as the a significant value of the DW statistic could also come from omitted variables, incorrect functional form, or dynamic misspecification.

dimitriy
  • 31,081
  • 5
  • 63
  • 138
  • Thank you! I suspected that the DW test showing was a simple coincidence. Because the distance between various regions was not specified and they were simply numbered as 1 to 15 without categorization. – Siyanda Oct 22 '12 at 18:32