Can someone please explain the dickey fuller test? ARe we just testing whether the coefficient is 0 and if it is zero we have nonstationarity and this will create a problem for the model? What are things I need to know about it? Thanks
here is what i know… or what i think i know… we want the time series model to be covaraince stationary…the b1 can not be 1…the problem is how do we test for this…so instead of testing directly if b1 = 1. d-f came up w/ the idea we sholud test if b1 - 1 =0… so when we test this if we reject the null, we conclude that b1 is not one and we are good to go… if we do not reject the null we conclude that b1-1 = 0, thus b1 = 1 and we have a unit root…
thanks Nikko i got it. I will finish the rest and probably have more questions. In addition to this can we add that if b1=1 the mean reverting will be infinite and so incorrect i.e. b0/(1-b1) = b0/(1-1) =infinite and so infinite?
if b1 =1 the there is no mean reverting level… it is undefined… 1 / 1-1 = 1/0 = undefined… thus no mean reverting level