Читать книгу Industrial Data Analytics for Diagnosis and Prognosis - Yong Chen - Страница 37

A random vector is a vector of random variables. Let X=(X1 X2 . . . Xp)T denote a random vector. The mean or expected value of a random vector is the vector of the mean values of each of its elements. The mean vector of X can be written as

where the univariate mean is defined as

where fi,(xi) is the probability density function of Xi if Xi is continuous and pi(Xi is the probability mass function of Xi if Xi is discrete. The µi is also called the population mean of Xi because it is the mean of Xi over all possible values in the population. Similarly, the mean vector µ is the population mean vector of X.

To further explain the relationship and difference between the population mean and the sample mean introduced in Section 2.2, we first consider a univariate random variable X and its population mean μ. Consider a random sample of observations from the population, say, X₁, X₂,…, Xn. The sample mean is a random variable because the observations X₁, X₂,…, Xn are all random variables with values varying from sample to sample. For example, let X represent the measured intensity of the current of a wafer produced by a semiconductor manufacturing process. Then we take a random sample of n = 10 wafers from this process and compute the sample mean of the measured intensities of the current and get the result x̄ = 1.02. Now we repeat this process, taking a second sample of n = 10 wafers from the same process and the resulting sample mean is 1.04. The sample means differ from sample to sample because they are random variables. Consequently, the sample mean, and any other function of the random observations, is a random variable. On the other hand, the population mean µ does not depend on the samples and is a (usually unknown) constant. When we take a sample with very large sample size n, the sample mean will be very close to the population mean µ with high probability. As the sample mean X̄ is a random variable, we can evaluate its mean and variance. It is easy to see that E(X̄) = µ and var (X̄) = σ²/n, where β² is the variance of X. An estimator of a parameter is called unbiased if its mean is equal to the true value of the parameter. X̄ is a commonly used estimator of µ because it is unbiased and has a smaller variance for a larger sample size n.

This concept can be extended to a p-dimensional random vector X with mean vector µ. Consider a random sample X₁, X₂,…, Xn from the population of X. The sample mean vector X̄ is a random vector with population mean E(X̄) = µ and population covariance matrix , where Σ is the population covariance matrix of X. The population covariance matrix is defined shortly. The sample mean vector X̄ is an unbiased estimator of the population mean vector μ.

The (population) covariance matrix of a random vector X is defined as

The ith diagonal element of Σ is the population variance of X_i:

The (j,k)th off-diagonal element of Σ is the population covariance between Xj and Xk:

where fjk(xj, xk) and pjk(xj, xk) are the joint density function and joint probability mass function, respectively, of Xj and Xk. The population covariance measures the linear association between the two random variables. It is clear that σi = σkj and the covariance matrix Σ is symmetric. The same as the sample covariance matrix, the population covariance matrix Σ is always positive semidefinite.

Similar to the population mean, the population variance and covariance can be estimated by the sample variance and covariance introduced in Section 2.2. The sample variance and covariance are both random variables, and are unbiased estimators of the population variance and covariance. Consequently, the sample covariance matrix S is an unbiased estimator of the population covariance matrix Σ, that is, E(S) = Σ.

As for the sample covariance, the value of the population covariance of two random variables depends on the scaling, possibly due to the difference of measuring unit of the variables. A scaling-independent measure of the degree of linear association between the random variables Xj and Xk is given by the population correlation:

It is clear that ρjk = ρkj. And the population correlation matrix of a random vector X is a symmetric matrix defined as

For univariate variables X and Y and a constant c, we have E(X + Y) = E(X) + E(Y) and E(cX) = cE(X). Similarly, for random vectors X and Y and a constant matrix C, it can be seen that

(3.1)

The covariance matrix of Z = CX is

(3.2)

The similarity of (3.2) and (2.10) is pretty clear. When C is a row vector c^T = (c₁, c₂,…, cp), CX = c^TX = c₁X₁ + … + cp Xp and

(3.3)

(3.4)

where μ and Σ are the mean vector and covariance matrix of X.

Let X₁ and X₂ denote two subvectors of X, i.e., . The mean vector and the covariance matrix of X can be partitioned as

(3.5)

(3.6)

where Σ₁₁ = cov(X₁) and Σ₂₂ = cov(X₂). The matrix Σ₁₂ contains the covariance of each component in X₁ and each component in X₂. Based on the symmetry of Σ, we have .