Efficient Estimation of Stochastic Parameters: A GLS Approach

Abstract

This thesis presents a novel rolling GLS-based model to improve the precision of time-varying parameter estimates in dynamic linear models. Through rigorous simulations, the rolling GLS model exhibits enhanced accuracy in scenarios with smaller sample sizes and maintains its efficacy when the normality assumption is relaxed, distinguishing it from traditional models like Kalman Filters. Furthermore, the thesis expands on the model to tackle more complex stochastic structures and validates its effectiveness through practical applications to real-world financial data, like inflation risk premium estimations. The research culminates in offering a robust tool for financial econometrics, enhancing the reliability of financial analyses and predictions.