均值、均方值、方差、均方差和协方差概念及其物理意义

提示：文章写完后，目录可以自动生成，如何生成可参考右边的帮助文档

1.均值（Mean）

A. Definition

Successive Type

The mean of Random Vriable $\mathbf{x}$ is
$$E(\mathbf{x})={\int }_{-\infty }^{\infty }xf(x)\text{d}x.$$

Discrete Type

Letting random variable $x$ with $x_i$ having $p_i$, we have
$$f(x)=\sum _i{p}_i\delta (x-x_i).$$ Substituting identical eqution ${\int }_{-\infty }^{\infty }x\delta (x-x_i)\text{d}x=x_i$ into the above equation, we have
$$E(x)=\sum _i{p}_i{x}_i,p_i=P(\mathbf{x}=x_i).$$

Conditional Mean

Let Conditional Probability Density (CPD) replace PDF $f(x)$, the conditional mean of random variable $\mathbf{x}$ with condition M is
$$E(\mathbf{x}|M)={\int }_{-\infty }^{\infty }xf(x|M)\text{d}x$$
For discrete random variable, the above equation is rewritten as $$E(\mathbf{x}|M)=\sum _i{x}_{i}P(\mathbf{x}=x_i|M).$$

B. Property

Mean can express the magnitude of direct current, denoted as $E(x)$.
For a Gaussian white noise signal, its mean is zero, and it only has alternating components.

Note that the square mean, ${E(x)}^2$, can express the power of direct components.

---

2. 均方值（Mean Square Value）

Mean Square Value can express the mean power of a signal, $$E(x^2)$$.
Mean power of a signal = Component of alternating current + Component of direct current

---

3. 方差（variance）

A. Definition

It can be expressed as variance or deviation or Var.

1）Successive Type

$$D(X)={\sigma }^2={\int }_{-\infty }^{\infty }(x-\mu {)}^{2}f(x)\text{d}x$$ where $\mu =E(X)$.
The above equation can be rewritten as $$D(X)=\int x^{2}f(x)dx-{\mu }^2=E\left(X^2\right)-[E(X){]}^2.$$

2）Discrete Type

$$D(X)=\sum _{i=1}^N{p}_i(x_i-\mu {)}^2=\sum _{i=1}^N(p_i{x}_i^2)-{\mu }^2.$$

In probability theory, variance can measure the deviation extent between random variables and mean.

Variance can describe the deviation extent of a signal and express the strength of alternating components, i.e., mean power of an alternating signal.

B. Property

• Letting $C$ be a constant, we have $D(C)=0$.
• Letting $X$ be a random variable, we have $D(CX)=C^{2}D(X)$, $D(X+C)=D(X)$.
• Letting $X,Y$ be random variables, we have $D(X\pm Y)=D(X)+D(Y)\pm 2Cov(X,Y)$, where C o v ( X , Y ) = E { [ X − E ( X ) ] [ Y − E ( Y ) ] } Cov (X,Y)=E\{[X-E(X)][Y-E(Y)]\} Cov(X,Y)=E{[X−E(X)][Y−E(Y)]} .

---

4. 标准差 （Standard Variance）

Standard Deviation is also called Root-Mean-Square Error (RMSE), denoted as $\sigma$, which can reflect the discrete degree of a dataset.

Note that Standard Variance can reflect the discrete degree between measuring data and true data, which can be a metric. The smaller the Standard Variance, the higher the accuracy.

---

5. 协方差 （Covariance）

In statistics and probability theory, Covariance can measure the total error between two variables.

$$\begin{array}{ll}Cov(X,Y)& =E\left[(X-E(X))(Y-E(Y))\right]\\ & =E\left[XY\right]-2E[X]E[Y]+E[X]E[Y]\\ & =E\left[XY\right]-E[X]E[Y]\end{array}$$

A. Meaning

Case 1:
如果两个变量的变化趋势一致，如果其中一个大于自身的期望值时另外一个也大于自身的期望值，那么两个变量之间的协方差就是正值；
Case 2:
如果两个变量的变化趋势相反，即其中一个变量大于自身的期望值时另外一个却小于自身的期望值，那么两个变量之间的协方差就是负值。即协方差具备描述X和Y相关程度的量的功能。

如果X与Y是统计独立的，那么二者之间的协方差就是0，因为两个独立的随机变量满足$E[XY]=E[X]E[Y]$。但是，反过来并不成立。

若两个随机变量X和Y相互独立，则$E[(X-E(X))(Y-E(Y))]=0$，因而若上述数学期望不为零，则X和Y必不是相互独立的，亦即它们之间存在着一定的关系。

B. The relationship between Variance and Covariance

$$D(X+Y)=D(X)+D(Y)+2Cov(X，Y)$$ $$D(X-Y)=D(X)+D(Y)-2Cov(X，Y)$$

协方差与期望值有如下关系：
$$Cov(X，Y)=E(XY)-E(X)E(Y)$$

C. 协方差的性质

• $Cov(X,Y)=Cov(Y,X)$；

• $Cov(aX,bY)=abCov(X,Y)$，($a$ and $b$ are constant)；

• $Cov(X_1+X_2,Y)=Cov(X_1,Y)+Cov(X_2,Y)$

D. 相关系数

相关系数是作为描述两者变量的相关程度的量。

$${\rho }_{XY}=\frac{\mathrm{Cov}(X,Y)}{\sqrt{D(X)}\sqrt{D(Y)}}$$

性质
设ρXY是随机变量X和Y的相关系数，则有

（1）$\mid {\rho }_{XY}\mid \le 1$；

（2）$\mid {\rho }_{XY}\mid =1$充分必要条件为 P { Y = a X + b } = 1 P\{Y=aX+b\}=1 P{Y=aX+b}=1,（$a$ and $b$为常数，$a\ne 0$）；

（3）若${\rho }_{XY}=0$，则称X与Y不线性相关。

---