MVTransfer Function Command

The MVTRANSFER_FUNCTION command estimates a multivariate transfer function model. It compares one target series with several explanatory time series, using ARIMA-filtered residuals and lagged relationships to identify the strongest delayed effects.

Example

MVTRANSFER_FUNCTION(LIST(GDP,IMPGS,EXPGS),24,7)

This example estimates a transfer function relationship between GDP, imports, and exports using up to 24 lags and selecting the best 7 explanatory variables.

Syntax

MVTRANSFER_FUNCTION(seriesList, lag, maxVars)

The first argument is a list of time series. The second argument is the maximum lag length. The third argument is the maximum number of variables to include in the best regression model.

What It Does

Each input series is first processed through an ARIMA model. The residuals are then lagged and used in a cumulative regression search. This helps separate the predictable time-series structure from the cross-series lag relationship.

The command also compares the ARIMA-filtered lag model with raw lag regressions, making it easier to judge whether the relationship is coming from true lagged interaction or from shared trend and autocorrelation.

What It Returns

• Actual vs. Fitted: compares the observed target series with the model prediction.
• Autocorrelation Analysis: checks whether the final regression residuals still contain serial correlation.
• Model: reports the selected regression coefficients and diagnostics.
• ARIMA Panels: shows the ARIMA model used for each input series.

Interpretation

A useful model should fit the target series while leaving residuals with low autocorrelation. Strong selected lag terms suggest that one variable helps explain delayed movement in another after each series has been filtered for its own internal dynamics.

This command is useful when several macroeconomic or market variables may affect a target variable over time, but the timing and importance of those effects are not obvious.

Typical Use Cases

• GDP response to imports, exports, rates, or policy variables
• Inflation response to money supply, energy, wages, or output
• Market response to macroeconomic indicators
• Trade-flow and tariff analysis
• Testing delayed cross-series relationships after ARIMA filtering

Notes

The command aligns all input series to the same date range before modeling. The lag parameter controls how far back the model searches for delayed effects, while maxVars limits the number of selected lagged variables.