Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 17 views

    Normal Distribution

    Uploaded by

    Prashanth9999
    This document defines normal random variables and Gaussian vectors. It states that a normal random variable X is characterized by its mean μ and variance σ^2 and has a probability density function involving these parameters. It then defines a Gaussian vector as a random vector whose linear combinations are normal random variables. A Gaussian vector X is characterized by its mean vector μ and covariance matrix Q. The document provides properties of normal random variables and Gaussian vectors, such as their characteristic functions and how independent normal random variables and Gaussian vectors combine.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Normal Distribution

    Uploaded by

    Prashanth9999
    17 views2 pages
    This document defines normal random variables and Gaussian vectors. It states that a normal random variable X is characterized by its mean μ and variance σ^2 and has a probability density function involving these parameters. It then defines a Gaussian vector as a random vector whose linear combinations are normal random variables. A Gaussian vector X is characterized by its mean vector μ and covariance matrix Q. The document provides properties of normal random variables and Gaussian vectors, such as their characteristic functions and how independent normal random variables and Gaussian vectors combine.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document defines normal random variables and Gaussian vectors. It states that a normal random variable X is characterized by its mean μ and variance σ^2 and has a probability density function involving these parameters. It then defines a Gaussian vector as a random vector whose linear combinations are normal random variables. A Gaussian vector X is characterized by its mean vector μ and covariance matrix Q. The document provides properties of normal random variables and Gaussian vectors, such as their characteristic functions and how independent normal random variables and Gaussian vectors combine.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    17 views2 pages

    Normal Distribution

    Uploaded by

    Prashanth9999
    This document defines normal random variables and Gaussian vectors. It states that a normal random variable X is characterized by its mean μ and variance σ^2 and has a probability density function involving these parameters. It then defines a Gaussian vector as a random vector whose linear combinations are normal random variables. A Gaussian vector X is characterized by its mean vector μ and covariance matrix Q. The document provides properties of normal random variables and Gaussian vectors, such as their characteristic functions and how independent normal random variables and Gaussian vectors combine.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    NORMAL RANDOM VARIABLES AND GAUSSIAN VECTORS

    1. Normal Random Variables


    Recall that a random variable X with values in R is Normal (or Gaussian) with parameters
    (, 2 ), R, 2 > 0, if its density with respect to the Lebesgue measure is given by:
    1
    2
    2
    e(x) /2 , < x < .
    f (x) =
    2
    Such a distribution is usually denoted by N (, 2 ), and we write X N (, 2 ). We can
    also include the case of a degenerate Normal random variable with distribution N (, 0). This
    is simply the Dirac distribution of a random variable which is almost surely equal to .
    The mean and the variance of a Normal random variable X are given by:
    Var(X) = 2 .

    E[X] = ,

    The characteristic function of a such a Normal random variable is:


    
    
     iX 
    2 2
    E e
    = exp i , R,
    2
    and the Laplace transform is:
    
    
     zX 
    z2 2
    E e
    = exp z + ,
    2

    z C.

    Note that of course the characteristic function follows immediately, when the Laplace transform is known.
    An important property of Normal random variables is that if X N (1 , 12 ), and Y
    N (2 , 22 ), and X and Y are independent, then X + Y N (1 + 2 , 12 + 22 ). This follows
    immediately from the form of the characteristic function.
    Proposition 1.1. Let (Xn ) be a sequence of Normal random variables, such that Xn
    N (n , n2 ), for all n. Suppose that Xn converges in distribution to X. Then, it follows that
    X is also Normal with mean = limn n , and variance 2 = limn n2 .
    2. Gaussian Vectors
    Definition 2.1. An Rn -valued
    P random variable X = (X1 , . . . , Xn ) is a Gaussian vector, if
    for any u Rn , < u, X >= ni=1 ui Xi is a (one-dimensional) Normal random variable.
    If X is a Gaussian vector, there exists Rn , and a quadratic positive form qX on Rn ,
    such that for all u Rn ,
    E[< u, X >] =< u, >,
    Var(< u, X >) = qX (u).
    Date: September 27, 2009.
    1

    In fact, if (e1 ,P
    . . . , en ) is the canonical orthonormal basis of Rn , and we write X =
    we have = ni=1 E[Xi ]ei =: E[X], and
    n
    X
    qX (u) =
    uj uk Cov(Xi , Xj ),

    Pn

    i=1

    Xi ei ,

    i,j=1

    where Q := Cov(Xi , Xj ) = E[(Xi E[Xi ])(Xj E[Xj ])].


    Since < u, X > N (< u, >, qX (u)), it is now easy to deduce the characteristic function
    of X:
    
    
    1
    X (u) = E[exp(i < u, X >)] = exp i < u, > qX (u) , u Rn .
    2
    Note that in fact, the characteristic function uniquely characterizes the distribution of X.
    Proposition 2.2. Two components Xl and Xk of a Gaussian vector are independent if and
    only if they are uncorrelated (ie. Qlk = Qkl = 0). In particular, the components X1 , . . . , Xn
    are independent if and only if the covariance matrix Q is diagonal.
    The following theorem is also useful:
    Theorem 2.3. Let X be and Rn -valued Gaussian vector and let Y an Rm -valued Gaussian
    vector. Assume that X and Y are independent. Then, Z = (X, Y ) is an Rn+m -valued
    Gaussian vector.
    The characteristic function of Z is Z (u) = X (w)Y (v), where u = (w, v), w Rn and
    v Rm .
    In fact, all Gaussian vectors arise as linear transformations of vectors of independent
    centered Normal random variables as the next theorem shows:
    Theorem 2.4. Let X be an Rn -valued Gaussian vector with mean vector . Then, there
    exist n independent Normal random variables Y1 , . . . , Yn , with Yj N (0, j ), j 0, and
    an orthogonal matrix A, such that X = + AY .
    Note that A is such that A QA = diag(1 , . . . , n ). Hence, the j s are the eigenvalues of
    the covariance matrix of X. Also, it may be possible, that some of the j s are zero, that
    is, Yj = 0 almost surely. So, the number of independent normal random variables needed to
    determine a Gaussian vector in Rn may be strictly less than n. In the case, where all the
    j s are positive, it is possible to give the density of X with respect to the Lebesgue measure
    on Rn . Namely, the density is given by:
    
    
    1
    1
    T 1
    p
    f (x) =
    exp (x ) Q (x ) .
    2
    (2)d/2 det(Q)

    You might also like