ARBIMETRA LAB
Moving Average Model (MA)
Models a series using past forecast errors.
What it does
Models a series using past forecast errors.
Use it for stationary series where the ACF cuts off after a few lags.
Data you need
- One series
- Order q
Supported data: Time series. Level: Intermediate.
Formula and method
yₜ = c + εₜ + Σθᵢεₜ₋ᵢ
Current values depend on present and past shocks, which cannot be observed. The Kalman filter reconstructs them, so the exact likelihood can be evaluated and maximised over the MA coefficients.
Assumptions
- The series is stationary after any differencing
- Errors are uncorrelated with constant variance
Input checks
- Order up to 12
- Unbroken series
How to read the result
Forecasts return to the mean after q steps.
Worked example
Illustrative run on US quarterly macro 1959-2009. These figures are the actual output of this model on that example, and they say nothing about your own data.
- AIC = 993.9967, BIC = 1,003.94.
- Next-period forecast: 4.0103, 95% interval -1.3917 to 9.4123.
- Last-period forecast (2011Q3): 3.9542, 95% interval -2.0841 to 9.9925.
- The residuals still show autocorrelation, so the model has not captured all the dynamics.
- Forecast intervals widen with the horizon: uncertainty grows the further ahead you look. A forecast is an estimate, not a prediction of what will happen.
Limitations
- Results depend on the sample period
Numerical validation
Validated 16 of 16 automated checks passed, including numerical results, invalid input, missing values, small samples, constant variables, charts and the interpretation text.
| Check | Reference | Result | Abs. error | Tolerance | Source |
|---|---|---|---|---|---|
| Likelihood against exact covariance formula | -299.65122087 | -299.65122087 | 1.1e-13 | 1e-6 | Exact Gaussian likelihood |
| Reported estimate matches a direct fit | 4.75233796 | 4.75233796 | 0.0e+0 | 1e-6 | statsmodels ARIMA direct call |
Computed with python 3.13.2, numpy 2.2.5, scipy 1.14.1, pandas 2.3.3, statsmodels 0.14.4, scikit-learn 1.7.0, in your browser. A passing check confirms the listed comparison only. It does not make any result suitable for your data, and it is not a claim of equivalence with other statistical software.
Questions
Why do MA forecasts become flat?
Beyond q steps ahead there are no known past errors to use, so the forecast equals the mean.
ARBIMETRA LAB is an educational and research tool. Results should be checked by the researcher and are not professional advice.
