Multivariate time series analysis using vars package with simplified IRF, FEVD, and correlation analysis for 2-10 variables
VAR model with type="both" (constant and trend), default lag order 2 or user-specified.
Simplified IRF analysis between specified variables with confidence bands.
Forecasts for all variables with upper/lower confidence intervals.
Provide a dataset with 2-10 numeric variables as an array. Optional date_column for time indexing.
Fits VAR model with both constant and trend, provides AIC/BIC information criteria, calculates correlation matrix between variables, generates forecasts for specified periods, and includes simplified FEVD (Forecast Error Variance Decomposition) analysis.
From stationarity to interpretable dynamics
Convert to time series object, set frequency (monthly default), handle JSON format.
Apply VAR with specified lag order, extract coefficients and model fit statistics.
Generate forecasts, calculate IRF if specified, create simplified FEVD and diagnostics.
Simplified VAR implementation for practical multivariate time series analysis—handles 2-10 interconnected variables efficiently.
The implementation uses the vars package, provides simplified placeholder values for complex statistics (Granger causality, stability checks), includes correlation matrix between all variables, generates forecasts with confidence intervals for all series, and offers optional IRF analysis between specified variables. ACF/PACF data and Portmanteau test results are simplified for stability.
Note: Uses VAR type="both" (constant and trend). Default lag order is 2. Monthly frequency assumed if date column provided. Some diagnostic values are simplified placeholders.