Dickey-Fuller test

To which extend do you think we need to know the Dickey-Fuller test ??? The LOS goes “discuss the steps of the unit root test for nonstationarity, and explain the relation of the test to autoregressive time series model” The think that bothers me is that we don’t directly test for b1=1 , but we pass by some g value, sounds fair enought but that’s one more detail to remember on exam day. Do you think that knowing that we (ultimately) test for b1=1 and that if we fail to reject the null (i.e. we conclude that b1=1), we’ll conclude that the series has a unit root, and hence is not covariance stationnary would eb enought???

I know it is juvenile but I can’t help but to chuckle at the test name Wikipedia has a nice summary. http://en.wikipedia.org/wiki/Dickey-Fuller_test http://en.wikipedia.org/wiki/Augmented_Dickey-Fuller_test You reject the null hypothesis that b1=1 (or unit root is present or non-stationary) if the test statistic is less than critical value If fail to reject, then it has a unit root except such claim is not very strong.

C’mon I’ve been chuckling at that name for years. Dickey-Fuller? What were they thinking?

my take on it is that you will be given the modified critical t-values in the exam. Main thing is to know what to do with it - ie if test-stat > DF critical val: --> reject null --> series is covariance stationary --> model good. Also, looks like they will provide Engle-Grainger ciritical vals for co-integration test of error terms using Dickey-Fuller as well. Main thing is to know what to do with it. Same test rule applies as above.

could have been better - they could have called it Fuller-Dickey ?!?!?