自迴歸隨機波動的向量自迴歸模型之馬蹄鐵估計

Abstract

在這篇研究中，我們會給出一個包含向量自迴歸模型、自迴歸隨機波動、馬蹄鐵先驗分佈的模型，再使用變分貝氏方法估計模型中的變數服從的分佈，像是常態分佈或是逆伽瑪分佈這些比較常見以及期望值容易計算的分佈，最後再使用迭代的方式找到每個變數的估計值。

In this thesis, we propose an approach to find an estimate of the variables in a model that combines vector autoregression (VAR), log-normal autoregressive stochastic volatility (ARSV) and the horseshoe prior. Using variational Bayes (VB) method, we can show an approximated distribution to each variable such as normal or inverse gamma distribution which are well-known and the expectation is easy to be obtained, which allows us to use iteration to estimate each variable.