How do we determine whether a regression is covariance stationary, given residuals from a AR time series … 1. Find T stat by dividing residuals with 1/sqrt(T) 2. Using T table to find if any residual is significant… 3. If so Auto correlation exists …We have to add lagged error terms etc.
use Dickey-Fuller to test for covariance stationary Xt-Xt-1=b0+(b1-1)Xt-1 +e if b1-1 is statistically significantly equal to 0, the model has unit root, meaning that it is not covariance stationary.
GetSetGo Wrote: ------------------------------------------------------- > How do we determine whether a regression is > covariance stationary, given residuals from a AR > time series … > > 1. Find T stat by dividing residuals with > 1/sqrt(T) > 2. Using T table to find if any residual is > significant… > 3. If so Auto correlation exists …We have to add > lagged error terms etc. Thats serial correlation for an AR. For stationarity, do what missjovi said, but to me its easier to say if you fail to reject, you have a unit root, if you reject, there is no unit root problem.
I write this as (x)t - (x)t-1 = b + g(x)t-1 + e. If g=0, then then you have a unit root and it is not stationary. Go on Dickey fuller, tell me if this shit is stationary or not! I’m done. Good luck tomorrow fellas.