多变量分布矩
-
二变量高斯分布:
\begin{equation}
f(u,v)=\frac{1}{2\pi\sigma_1\sigma_2\sqrt{1-\rho^2}}\exp\left(-\frac{1}{2(1-\rho^2)}\left[\frac{(u-\mu_1)^2}{\sigma_1^2}-\frac{2\rho(u-\mu_1)(v-\mu_2)}{\sigma_1\sigma_2}+\frac{(v-\mu_2)^2}{\sigma_2^2}\right]\right)
\end{equation}
MGF为:
\begin{equation}
m_{i,j}=\exp\left(i\mu_1+j\mu_2+0.5(\sigma_1^2i^2+2\rho\sigma_1\sigma_2ij+\sigma_2^2j^2)\right)
\end{equation}
纯矩计算方法为
\begin{equation}
\begin{split}
m_{0,0}&=1\\
m_{1,0}&=\mu\\
m_{2,0}&=\mu^2+\sigma^2\\
m_{3,0}&=\mu^3+3\mu\sigma^2\\
\end{split}
\end{equation}
假设$\mu_1=10,\mu_2=20,\sigma_1=\sigma_2=2,\rho=0.5$,有纯矩:
\begin{split}
m_{0,0}&=1\\
m_{1,0}&=10\\
m_{2,0}&=104\\
m_{3,0}&=1120\\
m_{0,1}&=20\\
m_{0,2}&=404\\
m_{0,3}&=8240\\
\end{split}
同时有混合矩
\begin{equation}
m_{i,j}=\exp\left(36\right)
\end{equation}
不行,混合矩计算方法不对