中文摘要：

以四只不同行业的热门股票为例，选用多元t分布拟合具有尖峰厚尾性的日收益率数据，运用均值一方差模型和最大化夏普比率找出其最优投资比例，计算最优投资组合在不同置信水平下的风险价值、期望损失和中位数损失．数据分析表明：期望损失和中位数损失均可以弥补风险价值的不足，且中位数损失比期望损失更稳健．利用三个尾部风险指标值可以为投资者控制风险提供多方位的参照，以达到防范风险减少损失的目的．

英文摘要：

The multivariate t distribution was employed to fit four hot stocks daily returns with peak and fat tails from different industries. The mean-variance model with maximizing the Sharpe Ratio was applied to find their optimal investment ratios. Then value at risk, expected shortfall and median shortfall were calculated under the different confidence levels for the optimal portfolio. The data analysis shows that both expected shortfall and median shortfall can make up the inadequacy of value at risk, and the performance of median shortfall is more robust than expected shortfall. Three tail risk indicators were introduced to provide investors more references to avoid risk and lessen loss.