TVP-VAR-SV Model

Overview

This repository contains MATLAB implementations for estimating and analyzing Time-Varying Parameter Vector Autoregressive (TVP-VAR) models with Stochastic Volatility (SV). These models are crucial for capturing dynamic relationships in time series data, allowing for changes in parameters and volatilities over time.

The repository includes scripts for:

• Model estimation using Bayesian techniques
• Impulse response function (IRF) estimation
• Volatility spillover analysis
• Parameter visualization
• Data preprocessing (log returns calculation, data import, etc.)

Key Features

• Bayesian Estimation of TVP-VAR-SV Models
• Dynamic Spillover Analysis
• Impulse Response Function Computation
• Time-Varying Parameter Visualization
• Log Returns Calculation for Financial Data

Repository Structure

├── data/               # Sample datasets for testing
├── scripts/            # MATLAB scripts for estimation and analysis
│   ├── estimateVARSV.m # Core function for TVP-VAR-SV estimation
│   ├── fitTVPVAR.m     # Bayesian fitting of TVP-VAR models
│   ├── netSpillovers.m # Spillover calculation function
│   ├── parameterPlots.m# Visualization of time-varying parameters
│   ├── impulseResponseEstimation.m # IRF computation
│   ├── calculateLogReturns.m # Data preprocessing script
│   ├── readData.m      # Function to import dataset
│   ├── rollingSamplePlots.m # Rolling window parameter analysis
├── results/            # Output files from model estimation
├── README.md           # Documentation

Installation and Usage

Prerequisites

• MATLAB R2022a or later
• Statistics and Machine Learning Toolbox (recommended)

Running the Model

1. Clone the repository:

   git clone https://github.com/wajoel/tvp-var-sv-model.git
   cd tvp-var-sv-model

2. Open MATLAB and navigate to the repository directory.
3. Load your dataset or use a sample dataset from data/.
4. Run estimateVARSV.m to estimate the model:

   results = estimateVARSV(data);

5. To compute impulse responses:

   irf = impulseResponseEstimation(results);

6. To analyze spillovers:

   spillovers = netSpillovers(results);

7. To visualize parameter evolution:

   parameterPlots(results);

Results

Below are outputs generated from the TVP-VAR-SV model:

References

• Nakajima, J. (2011). "Time-Varying Parameter VAR Model with Stochastic Volatility: An Overview."
• Joshua Chan & Eric Eisenstat (2018). "Bayesian Model Comparison for TVP-VAR-SV Models."

Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For any inquiries, please reach out via GitHub Issues or email the repository owner.