ARMAmodeller

ARMAmodeller is a software package designed for building and forecasting time series using autoregressive moving average (ARMA) models and their extensions. It provides a workflow for model identification, estimation, diagnostic checking, and forecasting within a unified environment.

Key features include automated identification of candidate AR and MA orders using ACF and PACF plots, information criteria (AIC, BIC) for model selection, and parameter estimation via maximum likelihood or conditional least squares. It supports AR, MA, ARMA, ARIMA, and SARIMA models, and can incorporate exogenous inputs (ARIMAX) and seasonal components.

The tool offers residual diagnostics such as Ljung-Box test, residual plots, and normality checks, along with visualization for fitted values and forecast intervals. It provides both a command-line interface and bindings for Python and R, enabling scripting and notebook workflows. Data can be imported from common formats, and missing values can be handled through imputation or pairwise deletion.

ARMAmodeller is used in finance, economics, engineering, and environmental sciences for short-term forecasting and scenario analysis. It emphasizes interpretability and reproducibility, exporting parameter estimates, diagnostics, and forecast distributions. Limitations include the need for stationarity, potential overfitting in small samples, and sensitivity to model misspecification.

Development and licensing: A project released as open-source software under a permissive license. Ongoing development integrates updates to estimation methods, diagnostics, and supported platforms.