本文採用RiskMetrics 模型與GARCH 模型及馬可夫轉換模型估算避險基金指數之風險值，並進一步以RiskMetrics 模型與GARCH 模型及馬可夫轉換模型所估出之風險值進行比較，用以探討何種模型有較佳的預測能力及績效，使投資大眾於面臨風險時，能正確的評估與控管，以避免承擔超過預期的損失。實證結果如下：1、由回溯測試的結果可知，RiskMetrics 模型與GARCH 模型及馬可夫轉換模型都能有效的估計風險值，風險控管能力均有一定的水準，其中又以馬可夫轉換模型在信心水準99%表現最佳。2、就資金使用效率的角度觀察，馬可夫轉換模型的表現為三模型中最優異的，推斷其原因為馬可夫模型採用馬可夫鍊做為狀態轉換的機制，相較於RiskMetrics 模型與GARCH 模型，更能夠考慮資料序列前後期狀態與相關訊息，進而對報酬分配有較精確的掌握。3、RiskMetrics 模型與GARCH 模型及馬可夫轉換模型進行回溯測驗及資金使用效率測驗時發現，穿透次數與均方根誤差存在有抵換關係。

This paper investigates the Value-at-Risk (VaR) of returns on hedge fund index using RiskMetrics model, GARCH model and Markov Switching Model. Furthermore, we compare with Valu-at-Risk (VaR) by RiskMetrics model, GARCH model and Markov Switching Model. The purpose is to find out which of three models has better prediction and performance for investors to evaluate and to take control in order to avoid unexpected lost while minimizing damage. The result of this study shows the following: (1)The back-test shows that the RiskMetrics model, the GARCH model and the Markov Switching Model can estimate Valu-at-Risk (VaR) effectively which proves that the ability to control risk is at good standard. Besides, the empirical results show Markov Switching Model can capture the distribution better than others in 99% confidence level under the back-test.(2)According to the efficiency of capital usage, the Markov Switching Model performs better than either the GARCH model or the RiskMetrics model. We infer that the Markov Switching Model can capture the distribution well resulting from it adopts the transformation mechanism of Markov chain. The Markov chain contains more relative information of time serial data than other models do. (3)All three models have the trade off between the back-test and efficiency of capital usage effectively.1