In 2013 morning Mock, the Durbin Watson statistic is used in a time series analysis. From what I remember, the Durbin Watson test is not to be used in time series and only for cross-sectional analysis. Can anyone clarify this?
Hi you are right you’re not supposed to use the DW statistics for testing for serial correlation in timeseries- they might have thrown it in there as a trick. To test for serial correlation I believe you use the regular t-test for a time series-
I believe Durbin Watson is ok for trend models but not ok for autoregression.
^I’ll second this, OK for time series but not for any AR-models.
I agree with CFA_shareholder. I was also under the assumption that DW test can not be applied to Time Series models.
Hi - apologies I want to clarify- and corroborate with Iconoclasts and MightyB - Use the DW statistic for Linear regression & Linear and Logn=linear Trend time series models - BUT NOT FOR AR MODELS. Sorry!
Right, the only time you want to use a DW test is when the data are time series… just not when you’ve used an AR model- applying DW to cross sectional data is meaningless as the data are in no particular temporal order