Panel VAR API Reference

Module

Import: from panelbox.var import PanelVAR, PanelVECM, PanelVARData Source: panelbox/var/

Overview

The VAR module implements Panel Vector Autoregression models for analyzing dynamic interdependencies among multiple variables across entities:

Class	Description	Reference
PanelVARData	Data container with lag structure	—
PanelVAR	Panel VAR estimation (OLS/GMM)	Holtz-Eakin, Newey & Rosen (1988)
PanelVECM	Panel Vector Error Correction Model	Johansen (1991)
CointegrationRankTest	Cointegration rank selection	—

Key features: lag order selection, impulse response functions (IRF), forecast error variance decomposition (FEVD), Granger causality tests, and forecasting.

Classes

PanelVARData

Data container for Panel VAR models. Handles lag construction, transformation, and variable management.

Constructor

PanelVARData(
    data: pd.DataFrame,
    endog_vars: list[str],
    entity_col: str,
    time_col: str,
    exog_vars: list[str] | None = None,
    lags: int = 1,
    trend: Literal["none", "constant", "trend", "both"] = "constant",
    dropna: Literal["any", "equation"] = "any",
)

Parameter	Type	Default	Description
data	pd.DataFrame	required	Panel data
endog_vars	list[str]	required	Endogenous variables (VAR system)
entity_col	str	required	Entity column
time_col	str	required	Time column
exog_vars	list[str] \| None	None	Exogenous variables
lags	int	1	Number of lags
trend	str	"constant"	Trend specification: "none", "constant", "trend", "both"
dropna	str	"any"	How to handle missing values

Example

from panelbox.var import PanelVARData

var_data = PanelVARData(
    data=df,
    endog_vars=["gdp", "inflation", "interest_rate"],
    entity_col="country",
    time_col="year",
    lags=2,
    trend="constant",
)

---

PanelVAR

Panel Vector Autoregression model. Supports OLS and GMM estimation with comprehensive analysis tools.

Constructor

PanelVAR(data: PanelVARData)

Parameter	Type	Description
data	PanelVARData	Prepared VAR data container

Methods

.fit()

def fit(
    self,
    method: str = "ols",
    cov_type: str = "clustered",
    **cov_kwds,
) -> PanelVARResult

Parameter	Type	Default	Description
method	str	"ols"	Estimation method: "ols" or "gmm"
cov_type	str	"clustered"	Covariance type

.select_lag_order()

def select_lag_order(
    self,
    max_lags: int,
    criteria: list[str] = ["aic", "bic", "hqic"],
) -> LagOrderResult

Select optimal lag order using information criteria.

.impulse_response()

Compute impulse response functions (orthogonalized or generalized).

.forecast_error_variance_decomposition()

Compute FEVD showing each variable's contribution to forecast error variance.

.granger_causality()

Test Granger causality between variable pairs.

.forecast()

Generate multi-step-ahead forecasts with confidence intervals.

Example

from panelbox.var import PanelVARData, PanelVAR

# Prepare data
var_data = PanelVARData(
    df, endog_vars=["gdp", "inflation", "interest_rate"],
    entity_col="country", time_col="year", lags=2,
)

# Estimate
model = PanelVAR(var_data)
result = model.fit(method="ols", cov_type="clustered")
print(result.summary())

# Lag order selection
lag_result = model.select_lag_order(max_lags=8)
print(lag_result.summary())

# Granger causality
gc = result.granger_causality("gdp", "inflation")
print(f"GDP -> Inflation: F={gc.statistic:.3f}, p={gc.pvalue:.3f}")

# Impulse response
irf = result.impulse_response(periods=20, method="cholesky")

# Forecast
forecast = result.forecast(periods=5, confidence=0.95)

---

PanelVECM

Panel Vector Error Correction Model for cointegrated panel data.

Constructor

PanelVECM(data: PanelVARData)

Methods

.fit()

def fit(
    self,
    coint_rank: int,
    lags: int | None = None,
) -> PanelVECMResult

Parameter	Type	Default	Description
coint_rank	int	required	Cointegration rank (number of cointegrating vectors)
lags	int \| None	None	Lag order (uses data's lags if None)

Example

from panelbox.var import PanelVARData, PanelVECM, CointegrationRankTest

var_data = PanelVARData(
    df, endog_vars=["consumption", "income", "wealth"],
    entity_col="household", time_col="quarter", lags=2,
)

# Test for cointegration rank
rank_test = CointegrationRankTest(var_data)
rank_result = rank_test.test()
print(rank_result.summary())

# Estimate VECM with selected rank
vecm = PanelVECM(var_data)
result = vecm.fit(coint_rank=rank_result.selected_rank)
print(result.summary())

---

CointegrationRankTest

Test for cointegration rank in Panel VECM.

Constructor

CointegrationRankTest(
    data: PanelVARData,
    max_rank: int | None = None,
    deterministic: str = "c",
)

Parameter	Type	Default	Description
data	PanelVARData	required	VAR data container
max_rank	int \| None	None	Maximum rank to test
deterministic	str	"c"	Deterministic specification

---

Result Classes

PanelVARResult

Result container for Panel VAR estimation.

Key Attributes

Attribute	Type	Description
params	dict	Coefficient matrices per equation
cov_params	dict	Covariance matrices
resid	pd.DataFrame	Residuals
sigma_u	np.ndarray	Residual covariance matrix
nobs	int	Number of observations

Methods

• .summary() — Formatted results for all equations
• .granger_causality(cause, effect) — Test Granger causality
• .granger_causality_matrix() — All pairwise causality tests
• .instantaneous_causality(var1, var2) — Contemporaneous causality test
• .impulse_response(periods, method) — IRF computation
• .forecast_error_variance_decomposition(periods) — FEVD
• .forecast(periods, confidence) — Multi-step forecasts

LagOrderResult

Lag order selection results with information criteria.

Key Attributes

Attribute	Type	Description
aic	dict	AIC by lag order
bic	dict	BIC by lag order
hqic	dict	HQIC by lag order
selected_lag	int	Recommended lag order

ForecastResult

Multi-step forecast results with confidence intervals.

PanelVECMResult

VECM estimation results with cointegrating vectors and adjustment coefficients.

RankSelectionResult

Cointegration rank test results.

RankTestResult

Individual rank test result (dataclass).

Attribute	Type	Description
rank	int	Tested rank
test_stat	float	Test statistic
z_stat	float	Standardized statistic
p_value	float	p-value
test_type	str	Test type
critical_values	dict \| None	Critical values at standard levels

---

Visualization

plot_causality_network

from panelbox.var import plot_causality_network

fig = plot_causality_network(var_result)

Generate a network visualization of Granger causality relationships.

References

• Holtz-Eakin, D., Newey, W. & Rosen, H. (1988). "Estimating vector autoregressions with panel data." Econometrica, 56(6), 1371-1395.
• Love, I. & Zicchino, L. (2006). "Financial development and dynamic investment behavior." QREE, 46(2), 190-210.
• Abrigo, M. & Love, I. (2016). "Estimation of panel vector autoregression in Stata." Stata Journal, 16(3), 778-804.

See Also

• GMM API — GMM estimation methods
• Diagnostics — Cointegration tests (Kao, Pedroni, Westerlund)
• Tutorials: VAR — Step-by-step Panel VAR guide