When a AR model is fit using the R arima package, you get back the estimate and S.E. for all the coefficients and the "intercept". However, for a non-zero-mean process, the true intercept is really $\phi_0=\mu(1-\phi_1-\phi_2-...)$ where $\mu$ is the intercept given by the arima output.
How does one go from the SE for $\mu$ and $\phi_1,..$ to the SE for $\phi_0$ assuming that all the estimates are N RVs? I believe that the $\phi_0$ is a $\chi^2$ RV here but can't figure out to use that to determine the SE.
Even a hint appreciated.
