Bayesian Vector Autoregression
Course Project · Bayesian Statistics
Bayesian Statistics
VAR Models
Time Series
Applying Bayesian priors to stabilize VAR(1) estimation for short psychological time series, comparing three approaches with different types of prior information.
For a master’s course in Bayesian Statistics, I explored how Bayesian methods can improve vector autoregression (VAR(1)) models, which are often used to study dynamic processes such as mood and behavior over time. These models are powerful but can easily overfit when applied to the short time series common in psychology. To address this, I implemented and compared three Bayesian approaches that use different types of prior information to stabilize estimation. This project gave me valuable insights into the practical complications of applying Bayesian VAR models and how methodological choices shape the results.