I have an elementary question regarding the error term in MA (ARIMA)--
From where does this error term come from?
From what I understood from the question raised earlier in the following link: Moving-average model error terms , the error term is the not the residual of regression yt (t is the subscript) over its previous value y(t-1), y(t-2),... (Please validate my understanding by answering this above sentence in YES/NO) but then how does the model get the estimates of the coefficients of the error terms from the data. Here both the error term and the coefficients are unknown, so how does it get the coefficients?
I am definitely missing something. Please help.
