-
Notifications
You must be signed in to change notification settings - Fork 1
ind_normal_wishart_favar_onestep
Jaromír Beneš edited this page Mar 20, 2026
·
1 revision
The One-Step FAVAR with Independent Normal-Wishart Prior implements simultaneous estimation of factors and VAR parameters using equation-by-equation independent Normal-Wishart priors. This estimator jointly models factor extraction and VAR dynamics with flexible equation-specific prior specifications, providing theoretically rigorous inference while allowing heterogeneous treatment across different variables.
The Independent Normal-Wishart FAVAR One-Step specification features:
- Simultaneous factor extraction and VAR parameter estimation
- Full Bayesian treatment of factor uncertainty
- Equation-by-equation independent Normal-Wishart priors
- Joint posterior inference over factors and equation-specific parameters
- Flexible prior specification allowing heterogeneity across variables
This approach combines theoretical rigor of one-step estimation with flexibility of independent equation treatment.
-
Sigma: Method of calculating priors on covariance matrix
- Values: "ar" (autoregressive) or "eye" (identity-based)
- Default: "ar"
- Controls how the prior covariance matrix is constructed
-
Autoregression: Prior on first-order autoregression
- Range: [0, 1]
- Default: 0.8
- Controls the belief in random walk behavior for own lags
-
Lambda1: Overall tightness of priors
- Range: (0, ∞)
- Default: 0.1
- Smaller values impose tighter priors (more shrinkage)
-
Lambda2: Variable weighting
- Range: (0, ∞)
- Default: 0.5
- Controls relative weights between own-variable and cross-variable lags
-
Lambda3: Lag decay
- Range: (0, ∞)
- Default: 1
- Controls how quickly prior importance decays with lag length
-
Lambda4: Exogenous variable tightness
- Range: (0, ∞)
- Default: 100
- Controls prior tightness on exogenous variable coefficients
-
Lambda5: Block exogeneity shrinkage
- Range: (0, ∞)
- Default: 0.001
- Controls shrinkage when block exogeneity is imposed
-
Exogenous: Priors on exogenous variables flag
- Default: false
- Enables specialized priors for exogenous variables
-
BlockExogenous: Block exogeneity flag
- Default: false
- Imposes block exogeneity restrictions
-
Burnin: Number of burn-in draws
- Range: [0, ∞)
- Default: 5000
- Required burn-in for joint factor-parameter sampling
-
It: Total number of iterations
- Range: [1, ∞)
- Default: 20000
- Total MCMC iterations including burn-in
-
StabilityThreshold: Threshold for maximum eigenvalue magnitude
- Range: (0, ∞)
- Default: null (no stability check)
- Ensures VAR stability by rejecting draws with eigenvalues above threshold
-
MaxNumUnstableAttempts: Maximum number of unstable sampling attempts
- Range: [1, ∞)
- Default: 1000
- Maximum attempts to find stable draw before stopping
- Heterogeneous variable treatment with factor uncertainty: When different variables require different priors and factor uncertainty matters
- Research applications: When methodological rigor is needed with equation-specific flexibility
- Mixed variable types: When system contains very different types of variables (financial, real, etc.)
- Robustness with rigor: When combining robust equation treatment with proper uncertainty quantification
- Large datasets: When computational burden becomes excessive
- Uniform prior beliefs: When all variables should be treated identically
- Real-time applications: When speed is more important than precision
- Simple applications: When two-step approaches provide adequate results
- More computationally intensive than two-step independent approaches
- Provides proper uncertainty quantification for heterogeneous systems
- Requires careful monitoring of convergence for each equation
- Benefits from equation-specific initialization strategies
Measurement Equation
X_t = Λ F_t + R_t + u_t
Independent Transition Equations For each variable i:
y_{i,t} = α_i + Σ_j β_{ij} y_{j,t-1} + Σ_k γ_{ik} f_{k,t-1} + ε_{i,t}
All parameters {Λ, F_t, {β_i, σ²_i}} estimated jointly with proper uncertainty accounting.
For each equation i:
β_i | σ²_i ~ N(m_i, σ²_i Ω_i)
σ²_i ~ IG(s_i, ν_i)
F_t ~ N(μ_F, Σ_F)
Λ ~ N(M_Λ, Σ_Λ)
- Draw factors F_t given all parameters and data
- Draw loadings Λ given factors and data
- For each equation i: Draw coefficients β_i given factors and equation-specific variance
- For each equation i: Draw variance σ²_i given coefficients and data
- Speed: Moderate speed due to equation independence but joint factor sampling
- Memory: Higher requirements for storing equation-specific and factor draws
- Convergence: Mixed convergence properties - factors may converge slower
- Scalability: Limited by both number of factors and equations
- Factors require joint sampling across all equations
- Equation parameters can be sampled independently given factors
- May exhibit different mixing rates across equation and factor blocks
- Benefits from adaptive sampling schemes for factor blocks
- Factors estimated jointly but affect each equation potentially differently
- Equation-specific factor loadings allow heterogeneous responses
- Factor uncertainty properly propagated to all equation-specific results
- More realistic uncertainty bands acknowledging both sources of uncertainty
- Equations treated independently conditional on factors
- Cross-correlations captured through common factor structure
- Residual correlations not directly modeled
- More robust to equation-specific outliers or breaks
- Factor uncertainty: Properly accounts for factor estimation uncertainty in all equations
- Consistent inference: Maintains theoretical coherence of Bayesian framework
- Uncertainty quantification: Better uncertainty bands for equation-specific results
- Factor-equation interaction: Proper treatment of factor-parameter interaction
- Computational cost: More expensive than two-step but less than system-wide one-step
- Implementation complexity: Moderate complexity balancing independence and joint estimation
- Convergence monitoring: Requires monitoring both factor and equation convergence
- Initialization: Benefits from sophisticated initialization for both components
- Use two-step independent estimates for equation parameters
- Initialize factors using principal components
- Set equation-specific starting values based on individual characteristics
- Consider warm-up periods for factor convergence
- Monitor convergence separately for factors and each equation
- Check factor identification and rotation issues
- Assess equation-specific effective sample sizes
- Validate posterior stability across all parameter blocks
- Use efficient algorithms for factor sampling given independence structure
- Exploit equation independence for parallel sampling where possible
- Consider adaptive schemes for factor blocks
- Monitor computational efficiency for both components
- Use convergence diagnostics for factor block
- Assess convergence for each equation separately
- Check cross-block interaction and stability
- Validate consistent interpretation across equations
- Examine factor loading distributions across all equations
- Assess equation-specific parameter estimates and uncertainty
- Compare factor vs. equation contributions to forecasting
- Test robustness of factor interpretation across equations
- Compare forecasting accuracy with system and two-step approaches
- Evaluate equation-specific prediction intervals
- Assess density forecast calibration
- Test structural interpretation consistency
- Banking and insurance variables with different regulatory effects
- Asset classes with varying sensitivities to macroeconomic factors
- International finance with country-specific institutional differences
- Mixed-frequency financial and real variables
- Industry-specific responses to aggregate shocks
- Regional analysis with area-specific characteristics
- Labor market analysis with demographic heterogeneity
- Environmental economics with sector-specific regulations
- Monetary policy transmission across different financial markets
- Fiscal policy effects varying across economic sectors
- International policy spillovers with country-specific institutions
- Regulatory impact analysis across heterogeneous entities
- Combines theoretical rigor with equation-specific flexibility
- Proper factor uncertainty treatment with heterogeneous priors
- More robust than system-wide approaches to equation misspecification
- Maintains factor coherence while allowing equation independence
- Computationally more intensive than two-step approaches
- Complex convergence monitoring requirements
- May lose some efficiency from equation independence
- Implementation complexity balancing multiple objectives
- Common factors affect different variables through different channels
- Equation-specific responses reveal heterogeneous transmission mechanisms
- Factor uncertainty affects each variable's inference differently
- Can identify which variables are most sensitive to factor uncertainty
- Variables share common driving forces but with different sensitivities
- Individual equation dynamics can vary while maintaining factor coherence
- Allows for asymmetric responses to common shocks
- Reveals variable-specific adjustment patterns to factor movements
- Bernanke, B.S., J. Boivin, and P. Eliasz (2005). "Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) Approach"
- Lopes, H.F. and M. West (2004). "Bayesian Model Assessment in Factor Analysis"
- Litterman, R.B. (1986). "Forecasting with Bayesian Vector Autoregressions-Five Years of Experience"
- Aguilar, O. and M. West (2000). "Bayesian Dynamic Factor Models and Portfolio Allocation"