![水灾害防治中的多变量概率问题](https://wfqqreader-1252317822.image.myqcloud.com/cover/642/37204642/b_37204642.jpg)
3.3 三变量联合概率分布
3.3.1 三维联合概率分布函数及重现期
设X,Y,Z为具有相关关系的随机变量,其联合分布函数定义为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_58.jpg?sign=1739348635-CmAy2zAjqEYfnmHnsiry0Lei034aQhOM-0-9d66c1f7c119cc32dabf543f41609e92)
式中:x,y,z分别为变量X,Y,Z的取值;F(x,y,z)为三维联合不超过概率。
则至少有一个变量被超过的联合重现期为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_59.jpg?sign=1739348635-4gMbfadj0ceHOnSSdsFPcopHWECbhDWl-0-5654473eb80c3ed0435871fd4aef986b)
当变量X,Y,Z相互独立时,其联合概率分布为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_60.jpg?sign=1739348635-TpJSZOY5xYi1sETv9zYEZvKiDE6KjVLs-0-e42d5bf3f603a533ccf1572bb2e0c916)
则各变量相互独立时的联合重现期表示为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_61.jpg?sign=1739348635-intw9zbW3PyKid0FH3ts8j9hlYZ2XmtD-0-9ab79efaf8689194f999b326cf1dd57a)
3.3.2 三维条件分布函数及重现期
设X,Y,Z为具有相关关系的随机变量,其联合概率密度函数为f(x,y,z);fX(x),fY(y),fZ(z)分别为X,Y,Z的边际概率密度函数,则:
(1)在Z=z条件下,事件(X≤x和Y≤y)的联合概率分布函数及概率密度函数为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_62.jpg?sign=1739348635-I9CuG5sLx8IS2Mo38ijZpVfGjMEBZti9-0-6f084d11df79c51605e6f39e4e924297)
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_63.jpg?sign=1739348635-okxEYhgl6gP08wU1nEnoGDD9OZw4AafN-0-8f23f949d9c8d0730e6e6faf3f147292)
相应的条件重现期表示为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_64.jpg?sign=1739348635-GRM2v8BktpsPakGfPgHKVKFpQ5OG28vH-0-b032eb8ad48d0f0117db8b23776e1a3d)
如果X,Y,Z相互独立,则:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_65.jpg?sign=1739348635-QLbUsbdyk8Mr7KvfmHJs2Z5GcPMn8Gld-0-870a6065cb44f92495724bada6a34bdb)
(2)在Y=y,Z=z条件下,事件X≤x的联合概率分布函数及概率密度函数为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_66.jpg?sign=1739348635-25uvl75qgoIfhocXXtpXBbotZ3Ms9WZg-0-9480774aa0939e68aca3c8c4ef038fe9)
相应的条件重现期为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_67.jpg?sign=1739348635-0cR76Ug5NWdofaYzU5GZHxU3OPIokTyv-0-9d7a01bbee6e44c9fbc3ba9dcee06250)
如果X,Y,Z相互独立,则:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_68.jpg?sign=1739348635-btFgzkwjbN9WgUU6snrVhuS0VkUb89PK-0-61c7d785ab18ff4f3640ff009b958184)
(3)在Z≤z条件下,事件(X≤x和Y≤y)的联合概率分布函数为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_69.jpg?sign=1739348635-HtHFMvrkEhmb9T4QuQkNSLp5DF74vgAB-0-77ffbfdf6906af8b8ae49b108c0ecbcf)
相应的条件重现期为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_70.jpg?sign=1739348635-bARw3SFIDlSTgWuF6wRToHleBXusA5sS-0-09ad5e35580b35094e338329205653ef)
如果X,Y,Z相互独立,则:F(x,y|Z≤z)=FX(x)FY(y)。
(4)在Y≤y,Z≤z条件下,事件X≤x的联合概率分布函数为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_71.jpg?sign=1739348635-QZU9CzKkFp0DwtcsjlVETiwbymIWVCOV-0-dcc0e81e4e702d2f9d2bd4f7aeddd314)
相应的条件重现期为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_72.jpg?sign=1739348635-TrQnhZ2PUYKgbWGDJBZu2VHXq3clnBg6-0-fd7f2981b7c63a77fcef8b01f7cf6e65)
3.3.3 三维联合概率分布模型
当变量维数n≥3,多变量联合概率分布问题因其复杂性难以有明确的解析表达式,只有在各变量均属正态分布时,其联合分布函数才会有解析表达式。
设三维随机向量X=(X1,X2,X3)服从参量为(μ,∑)的三维正态分布,记作X~N(μ,∑),则其概率密度函数为:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_73.jpg?sign=1739348635-nYKaRg9B9wrRmcrneHwP0t7dQ8u4frXe-0-adbaff99e88b4d57793d9c872b268682)
式中:μ=EX为数学期望向量;∑=DX=(X-EX)(X-EX)T为方差矩阵,∑-1为∑的逆矩阵;(X-μ)T为(X-μ)的转置;det∑表示矩阵∑的行列式。
各参数表达式如下:
![img](https://epubservercos.yuewen.com/DF778D/19720710008529106/epubprivate/OEBPS/Images/txt003_74.jpg?sign=1739348635-SX4kCxOzDV8PZdsgyCMrIhGCDhooCawI-0-cd1804bccbd8a85beaf9b372b8c0b94f)
三维正态分布模型由于计算较为复杂,且需要对变量边际分布进行正态化转换会影响分析的准确性,因此,在实际中较少应用。