Time series modeling pt 3. We'll walk through how we modify that basic model from white noise into more and more complicated time series modeling incorporating what we've learned about autoregressive and moving average processes.
Madam you made so easy. Perhaps I have been searching for this content. Time series is presented as difficult subject but you have made it so easy through step by step approach.
Should the ARMA and ARIMA models not have a e_t-1 term, instead of two y_t-1 terms on the right hand side, for the moving average part of the model? Or am I perhaps not understanding something
in the formula for ARIMA, Did you mean: theta1 * E(t-1), instead of: theta1 * Y(t-1) ? Also, in the general formula for ARMA, do you mean: Y(t+1) = ... instead of: Y(t) = ... I dont see why we would include the error term E(t) if we're predicting for time t And how can we use the model to predict multiple steps into the future?
Yes, the first one is mistake and your revised version is correct. We still need an error term because it captures the uncertainty related to that time step. If you're interested in multi-step predictions checkout our videos on forecasting with these models: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-kyPg3jV4pJ8.html
Technically I think this would be an AR(3) model that has coefficients of zero for the Yt-1 and Yt-2 terms, but I think the clearest thing to do is to just describe it explicitly as an AR model with only a Yt-3 term or better yet including the equation.
