NC-VAR-Selector 10.2

NC-VAR-Selector is a specialized tool for time-series practitioners that streamlines the discovery of parsimonious, stable Vector Autoregression (VAR) models. It automates variable screening and lag selection, compares competing specifications with transparent criteria, and produces ready-to-share outputs so you can focus on interpretation and forecasting.

Key features:

• Automated search over variables and lag orders with ranked model suggestions
• Multiple selection criteria (e.g., AIC, BIC, HQC, out-of-sample error) with customizable tie-breakers
• Optional regularization for high-dimensional datasets
• Stationarity checks and basic transformations (differencing, detrending, scaling)
• Diagnostics for residual autocorrelation and model stability
• Flexible constraints to force-include or exclude variables and cap maximum lags
• Reproducible workflows with run logs and fixed seeds
• Exportable summaries, parameters, and comparison reports

Use cases:

• Macroeconomic and policy forecasting
• Financial time-series modeling (rates, spreads, returns)
• Energy load and demand forecasting
• Multivariate sensor and IoT analytics

Benefits:

• Faster, defensible model selection with fewer manual iterations
• Clear trade-offs between accuracy and parsimony
• Cleaner documentation for auditability and collaboration

The most popular version of NC-VAR-Selector among our users is 10.2. The name of the program executable file is ncvar.exe.