about DW in an AR(1) model , but I could not understand the answer. anybody have a hint? Thanks.
From a time series perspective, DW can be used to test serial correlation only for the trend model or for the log linear models. For eg. DW can be used to test serial correlation for the following models. y = b0 + b1t or ln y = b0 + b1t etc. But, Q5 in page 500 does not have one of the above models. The model is Q 5 is an Auto regressive model with one lag. yt = b0 + b1 * y(t-1). For this model, where the independent variable is a lagged term of the dependent varianble (AR(1)), we can not use DW to test for serial correlation. We need to use a t-statistic on the autocorrelation of the residuals. (See page 454) Paddy
thanks. i hope I will have good luck… 
psriniva Wrote: ------------------------------------------------------- > From a time series perspective, DW can be used to > test serial correlation only for the trend model > or for the log linear models. > For eg. DW can be used to test serial correlation > for the following models. > y = b0 + b1t or > ln y = b0 + b1t etc. > > But, Q5 in page 500 does not have one of the above > models. > The model is Q 5 is an Auto regressive model with > one lag. > yt = b0 + b1 * y(t-1). > > For this model, where the independent variable is > a lagged term of the dependent varianble (AR(1)), > we can not use DW to test for serial correlation. > > We need to use a t-statistic on the > autocorrelation of the residuals. (See page 454) > Paddy