confused on MA in time series


New Member
I have taught myself how to complete time series calculations but I an confused on one point. When calculating a MA1 model, what does the data consist of? I know that they are the past errors. So am I going to do a EWMA or EMA model and take the errors from that and use them as predictor variables for the time series models? I know that for an AR model I take past values but for a MA model (so that I can do a ARIMA or ARMA or plain MA model) I am confused on what values that I should use. I am not sure if I should just take the errors that I get from a simple EWMA or EMA model and use those values to perform an acceptable regression/time series model. Can anyone help me out with what I should use? Thank you


Fortran must die
If you are going to do ARIMA you just use the raw data and the p,d,q you decide on. I have never seen anyone take the errors and I am not sure exactly what you are suggesting in honesty. An MA is a pattern you identify in the data, you do not take the errors yourself and do anything. The ARIMA does. Note arima is not regression, if you want to do that you can try regression with arima errors or if you enjoy pain multivariate arima.