Moving Average Models: MA(q) Noise, Autocorrelation, and Forecasting

Summary

The document explains the moving average time-series model of order q, in which an observation is represented using a constant and a finite set of current and past white-noise shocks. This differs from an autoregressive model, which uses prior observations and can reflect the effects of older shocks indirectly. The order q determines how many past noise terms enter the model.

It describes the model’s constant mean and finite autocovariance and autocorrelation: dependence vanishes beyond lag q. It also notes parameter estimation by maximum likelihood or nonlinear least squares, and positions MA(q) as a component of ARMA models and a tool for forecasting short-run fluctuations. Applications mentioned include financial prices and exchange rates, but no empirical example or measured forecast accuracy is provided. Forecast quality depends on selecting the order and estimating parameters well, and the article’s displayed formulas are incomplete in places, so it serves as a conceptual overview rather than a full modeling procedure.

Key ideas

• An MA(q) model represents observations using a constant and a finite number of white-noise terms.
• The model’s autocorrelation cuts off after lag q.
• Parameters may be estimated with maximum likelihood or nonlinear least squares.
• MA models can be combined with autoregressive terms in ARMA models.
• Forecasting performance depends on model order and parameter estimates.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.