Closed azev77 closed 3 years ago
You can reparameterize that model to be a linear regression with AR(1) errors. See https://robjhyndman.com/hyndsight/arimax/ Note: this is not a Q&A site. It is intended to report bugs or request features for the forecast package. Please ask such questions on crossvalidated.com or stackoverflow.com
Suppose I want to estimate an AR1 w/ an x variable:
y_t = \mu + \rho * y_t-1 + \beta * x_t + \epsilon_t
how is this possible w/ auto.arima()?It appears from the example in fpp2 that
auto.arima(y, xreg=x)
only fits regression models w/ ARIMA errors