I have made this mistake now three times so perhaps someone can help me. How do I know that a model is a moving average model (be it ARMA or MA) ?
That may not be such a basic question. Moving average and ARMA is not even mentioned in Schweser! Or did I just get sent the dodgy version of the Study Notes? It’s not in the LOS. And CFA notes say most time series data best modeled with AR. CFA notes says ARMA can be very unstable and have severe limitations. Having said that my very limited understanding of distinguishing MA from AR is via autocorrelations. MA autocorrelations will drop to zero suddenly after p lags for a MA § model whereas AR autocorr will decrease gradually through the lags. ARMA has a both AR lag of dependent terms and MA error terms. So if there are lagged dependent terms and lagged error terms, it is probably a ARMA. Have you been doing questions on this topic?
Is the question about how to distinguish MA from AR models in data analysis? Very hard and way beyond the scope.
Hm, let me be a bit more specific. For instance; the model xt = b0 + b1 xt-1 + b2 xt-2 + b3 xt-3 + b4 xt-4 + åt How do I know if this is an AR, MA or ARMA? I’m pretty sure something like this is AR but is it MA too?
Sort-of, but in a way that isn’t clear there. Call that one AR. MA is when there is an averaging of some other process.