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

Multivariate Normal Distribution: 1 Random Vector

The document defines and describes the multivariate normal distribution. It states that a random vector X is a vector of random variables and can be either discrete or continuous. For a continuous random vector X, its distribution is defined by a joint probability density function f. The multivariate normal distribution is defined as the distribution of a random vector AX, where X is a vector of independent standard normal random variables. The distribution only depends on the matrix A through AAT and has a mean of μ and covariance matrix of Σ. A formula for the joint probability density function of a multivariate normal distribution with mean μ and covariance matrix Σ is also provided.

Uploaded by

Saravanan Mathi
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
57 views

Multivariate Normal Distribution: 1 Random Vector

The document defines and describes the multivariate normal distribution. It states that a random vector X is a vector of random variables and can be either discrete or continuous. For a continuous random vector X, its distribution is defined by a joint probability density function f. The multivariate normal distribution is defined as the distribution of a random vector AX, where X is a vector of independent standard normal random variables. The distribution only depends on the matrix A through AAT and has a mean of μ and covariance matrix of Σ. A formula for the joint probability density function of a multivariate normal distribution with mean μ and covariance matrix Σ is also provided.

Uploaded by

Saravanan Mathi
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Multivariate Normal Distribution

September 30, 2008

Random Vector

A random vector X = (X1 , x2 , , Xk )T is a vector of random variables.

1. Discrete Case

If X takes value on a nite or countable set (or each Xi is a discrete random variable),we say X is a discrete random vector. In this case, the distribution of X is driven by the joint probability function. p(t1 , t2 , , tk ) = P (X1 = t1 , , Xk = tk ). 2. Continuous Case

In this case, the distribution of X is driven by the joint probability density function f (x1 , , xk ). The joint density function f satises that for any measurable set A Rk . P (X A) =
A

f (x1 , , xk )dx.

We can also dene the joint cdf, F , of X


t1 tk

F (t1 , , tk ) = P (X1 t1 , , Xk tk ) =

f (x1 , , xk )dx1 dxk .

It is easy to see that f (x1 , , xk ) = 3. Moments E (X ) = (E (X1 ), , E (Xk ))T COV (X ) = E ((X EX )(X EX )T ) = E (XX T ) E (X )E (X )T . It can be seen that for any matrix A, COV (AX ) = ACOV (X )AT . The moment generating function of X is dened as (for t Rk ) MX (t) = E (et
T X )).

k F (x1 , , xk ). x1 xk

Multivariate Normal Distribution

Suppose X = (X1 , , Xk ) and Xi are i.i.d. standard normal random variables. Then it is obviously that E (X ) = (0, 0, , 0), COV (X ) = Ik . Then for a n dimensional vector and n k matrix A E ( + AX ) = , COV ( + AX ) = AAT . Denote AAT by , we have the following denition. Denition 1 The distribution of random vector AX is called a multivariate normal distribution with covariance matrix and is denoted by N (0, ). And the distribution of + AX is called a multivariate normal distribution with mean and covariance matrix , N (, ).

To make the denition valid, we need to verify that the distribution of AX depend on A only throuth AAT . We can use the moment generating function to do this.

Suppose the moment generateing function of X is M (t), we know that M (t) = et So the m.g.f. of AX is MAX (t) = E (et
T AX

T t/2

) = M (tT A) = et

T AAT t

This means the m.g.f. of AX depend on A only through AAT , so the distribution of AX only depends on AAT . Based on the denition, we can also calculate the joint pdf of N (, ) (when is full rank),
1 1 T 1 f (x) = ( )n (det||)1/2 e 2 (x) (x) 2

where x = (x1 , , xn )T is a n dimensional vector. We can also see that if Y follows N (, ) distribution then for any matrix B BY N (B, B B T ).

You might also like