Durbin Watson Statistic
Candlefocus EditorThe DW statistic runs from 0 to 4, with a value of 2.0 meaning there is no autocorrelation. Values that are lower than 2.0 suggest positive autocorrelation, while higher than 2.0 point to negative autocorrelation. Positive and negative autocorrelation refer to the linear relationship between residuals (in a normal distribution). For example, a positive autocorrelation would be when the residuals after a positive error tend to have a positive error the next time.
The DW statistic is used to test the randomness of the residuals. When dealing with autocorrelation, there are two primary tests that are used. The first is the autocorrelation coefficient, or ACF. This test focuses on the temporal nature of autocorrelation, and complies with a lagged-correlation pattern. The second is the Durbin Watson statistic, which is a more general test for autocorrelation.
The Durbin-Watson statistic is used to assess the linearity of the regression model by calculating the correlation between the residuals. In the case of a normal regression model, it is assumed the residuals will be randomly distributed. If the DW statistic is below 2.0, this assumption is violated, suggesting the linearity of the regression model is not strong enough. This can be due to a distortion of the data caused by autocorrelation as well as other possibilities.
In summary, the Durbin Watson statistic is a useful test for determining autocorrelation in regression models. It ranges from 0 to 4, with a value of 2.0 indicating no autocorrelation. Values below 2.0 indicate positive autocorrelation, while values above 2.0 indicate negative autocorrelation. The DW statistic is primarily used to assess the linearity of the regression model by calculating the correlation between the residuals against randomness. It is a valuable tool that can help technical analysis, which is used to assess security prices.