I am using a Durbin-Watson test as a method through which to test for autocorrelation in a time series. Said data forms the basis of an interrupted time series analysis. My question is whether the absence of detected autocorrelation in the DW test on a simple model (OLS) is sufficient to inform modelling thereafter (ie. should I undertake the DW test on other plausible model types or does the DW test on an OLS suffice)?
Even if the DW test on a simple OLS model fails to detect autocorrelation, it is generally recommended to carry out additional diagnostic tests for autocorrelation in time series analysis. The Ljung-Box test, the Breusch-Godfrey test, and the Durbin h test are also diagnostic tests. In addition, it is essential to observe the residuals for any patterns or trends that may indicate autocorrelation. Generally, it is ideal to conduct multiple diagnostic tests and evaluate the results collectively to ensure that there is no autocorrelation in your data.
A proper answer to your question would require a knowledge of the economic model or theory that you are estimating and of the data that you are using. The fact that your residuals are not autocorrelated is not sufficient to be satisfied with your model. You also may need to do other misspecification tests such as tests for heteroskedasticity, functional form, stability of coefficients (recursive residuals), etc. The DW test is in respect of 1st order autocorrelation only. As Dickson Utonga has suggested you should use a more general test. (Note that the DW test is not valid if you have a lagged dependent variable in your model)
Hi,
Just to mention (refer to wikipedia), "the Durbin–Watson statistic, while displayed by many regression analysis programs, is not applicable in certain situations. For instance, when lagged dependent variables are included in the explanatory variables, then it is inappropriate to use this test. Durbin's h-test or likelihood ratio tests, that are valid in large samples, should be used".
Hamid