Time Series - Serial Correlation

CFA Level II
#1

Hello,

A schweser answer has stated that because a model has regressed macro data against time, it is using a trend model and therefore we can detect serial correlation using durbin watson.

I was under the impression DW cannot be used in time series. Can someone please clarify when we can use DW and when we cannot in Time Series?

#2

Serial correlation appear to be more frequent with time series and one way to test serial correlation is the DW test so yes, you do need to use DW when time series.

#3

Hi Djegu,

Thanks but I am looking for a more specific answer.

For example, with an AR model DW cannot be used I dont think.

Can anyone please provide specifically when we can use DW and when we cannot in time series?

#4

sorry was confused

Then

Multiple Linear Regression : Yes

Trend Models : Yes

AR : as you said , no

#5

We can use the DW test in any model in which there is no lagged dependent variable as independent variables in the model.

#6

From what I understand, it’s not appropriate to use the DW test for testing serial correlation of an AR model. You would use the t-test where the standard error is 1/sqrt(T). T = number of observations, and the degrees of freedom is T-2