Moment Generating Function. STAT/MTHE 353: 5 Moment Generating Functions and Multivariate Normal Distribution

Similar documents
Chapter 7: Special Distributions

The Multivariate Normal Distribution. In this case according to our theorem

Recall that if X 1,...,X n are random variables with finite expectations, then. The X i can be continuous or discrete or of any other type.

BIOS 2083 Linear Models Abdus S. Wahed. Chapter 2 84

The Multivariate Gaussian Distribution

Notes on Random Vectors and Multivariate Normal

MATH 829: Introduction to Data Mining and Analysis Consistency of Linear Regression

General Random Variables

5. Random Vectors. probabilities. characteristic function. cross correlation, cross covariance. Gaussian random vectors. functions of random vectors

Random Vectors and Multivariate Normal Distributions

3. Probability and Statistics

Numerical Linear Algebra

Multivariate Gaussian Distribution. Auxiliary notes for Time Series Analysis SF2943. Spring 2013

Lecture 11. Multivariate Normal theory

2. Matrix Algebra and Random Vectors

BASICS OF PROBABILITY

1.12 Multivariate Random Variables

Continuous Random Variables

The Multivariate Normal Distribution 1

Lecture 15: Multivariate normal distributions

Section 8.1. Vector Notation

The Multivariate Normal Distribution. Copyright c 2012 Dan Nettleton (Iowa State University) Statistics / 36

Probability- the good parts version. I. Random variables and their distributions; continuous random variables.

EEL 5544 Noise in Linear Systems Lecture 30. X (s) = E [ e sx] f X (x)e sx dx. Moments can be found from the Laplace transform as

Lecture 14: Multivariate mgf s and chf s

ECE 650 Lecture 4. Intro to Estimation Theory Random Vectors. ECE 650 D. Van Alphen 1

MLES & Multivariate Normal Theory

Gaussian random variables inr n

8 - Continuous random vectors

Review of Probability Theory II

Chapter 5. The multivariate normal distribution. Probability Theory. Linear transformations. The mean vector and the covariance matrix

Multivariate Random Variable

Review of Linear Algebra

Chapter 5. Chapter 5 sections

Chapter 5 continued. Chapter 5 sections

Random Variables. Random variables. A numerically valued map X of an outcome ω from a sample space Ω to the real line R

Random Variables and Their Distributions

The Multivariate Normal Distribution. Copyright c 2012 Dan Nettleton (Iowa State University) Statistics / 36

18 Bivariate normal distribution I

Formulas for probability theory and linear models SF2941

STAT 450. Moment Generating Functions

ANOVA: Analysis of Variance - Part I

Next tool is Partial ACF; mathematical tools first. The Multivariate Normal Distribution. e z2 /2. f Z (z) = 1 2π. e z2 i /2

7. Introduction to Large Sample Theory

The Multivariate Normal Distribution 1

P (x). all other X j =x j. If X is a continuous random vector (see p.172), then the marginal distributions of X i are: f(x)dx 1 dx n

Multivariate Distributions (Hogg Chapter Two)

Matrix Algebra, Class Notes (part 2) by Hrishikesh D. Vinod Copyright 1998 by Prof. H. D. Vinod, Fordham University, New York. All rights reserved.

Elements of Asymptotic Theory. James L. Powell Department of Economics University of California, Berkeley

Partial Solutions for h4/2014s: Sampling Distributions

Multiple Random Variables

Def. The euclidian distance between two points x = (x 1,...,x p ) t and y = (y 1,...,y p ) t in the p-dimensional space R p is defined as

x. Figure 1: Examples of univariate Gaussian pdfs N (x; µ, σ 2 ).

YORK UNIVERSITY. Faculty of Science Department of Mathematics and Statistics MATH M Test #1. July 11, 2013 Solutions

MAS223 Statistical Inference and Modelling Exercises

An Inverse Problem for Two Spectra of Complex Finite Jacobi Matrices

Math Bootcamp An p-dimensional vector is p numbers put together. Written as. x 1 x =. x p

Random vectors X 1 X 2. Recall that a random vector X = is made up of, say, k. X k. random variables.

Elements of Asymptotic Theory. James L. Powell Department of Economics University of California, Berkeley

MIT Spring 2015

1 Solution to Problem 2.1

E[X n ]= dn dt n M X(t). ). What is the mgf? Solution. Found this the other day in the Kernel matching exercise: 1 M X (t) =

BMIR Lecture Series on Probability and Statistics Fall 2015 Discrete RVs

Chater Matrix Norms and Singular Value Decomosition Introduction In this lecture, we introduce the notion of a norm for matrices The singular value de

2 b 3 b 4. c c 2 c 3 c 4

Part IA Probability. Theorems. Based on lectures by R. Weber Notes taken by Dexter Chua. Lent 2015

ACM 116: Lectures 3 4

x 3y 2z = 6 1.2) 2x 4y 3z = 8 3x + 6y + 8z = 5 x + 3y 2z + 5t = 4 1.5) 2x + 8y z + 9t = 9 3x + 5y 12z + 17t = 7

1 Linear Algebra Problems

p. 6-1 Continuous Random Variables p. 6-2

Economics 620, Lecture 5: exp

Stat 5101 Notes: Algorithms (thru 2nd midterm)

Knowledge Discovery and Data Mining 1 (VO) ( )

STAT 100C: Linear models

1.1 Review of Probability Theory

From Lay, 5.4. If we always treat a matrix as defining a linear transformation, what role does diagonalisation play?

conditional cdf, conditional pdf, total probability theorem?

2 Functions of random variables

Improving AOR Method for a Class of Two-by-Two Linear Systems

Exam 2. Jeremy Morris. March 23, 2006

LECTURES 2-3 : Stochastic Processes, Autocorrelation function. Stationarity.

Linear Algebra, part 2 Eigenvalues, eigenvectors and least squares solutions

Matrix Algebra. Matrix Algebra. Chapter 8 - S&B

Introduction to Probability and Stocastic Processes - Part I

Probability Lecture III (August, 2006)

Econ 508B: Lecture 5

The Properties of Pure Diagonal Bilinear Models

The purpose of this section is to derive the asymptotic distribution of the Pearson chi-square statistic. k (n j np j ) 2. np j.

Random Vectors 1. STA442/2101 Fall See last slide for copyright information. 1 / 30

5.1 Consistency of least squares estimates. We begin with a few consistency results that stand on their own and do not depend on normality.

Probability and Distributions

5 Operations on Multiple Random Variables

Computational Methods CMSC/AMSC/MAPL 460. Eigenvalues and Eigenvectors. Ramani Duraiswami, Dept. of Computer Science

2.3. VECTOR SPACES 25

11 a 12 a 21 a 11 a 22 a 12 a 21. (C.11) A = The determinant of a product of two matrices is given by AB = A B 1 1 = (C.13) and similarly.

The Hasse Minkowski Theorem Lee Dicker University of Minnesota, REU Summer 2001

Towards understanding the Lorenz curve using the Uniform distribution. Chris J. Stephens. Newcastle City Council, Newcastle upon Tyne, UK

DISCRIMINANTS IN TOWERS

ECE Lecture #10 Overview

APPENDIX A. Background Mathematics. A.1 Linear Algebra. Vector algebra. Let x denote the n-dimensional column vector with components x 1 x 2.

Transcription:

Moment Generating Function STAT/MTHE 353: 5 Moment Generating Functions and Multivariate Normal Distribution T. Linder Queen s University Winter 07 Definition Let X (X,...,X n ) T be a random vector and t (t,...,t n ) T R n.themoment generating function (MGF) is defined by M X (t) E e tt X for all t for which the exectation exists (i.e., finite). Remarks: M X (t) E e P n i tixi For 0 (0,...,0) T,wehaveM X (0). If X is a discrete random variable with finitely many values, then M X (t) E e tt X is always finite for all t R n. We will always assume that the distribution of X is such that M X (t) is finite for all t ( t 0,t 0 ) n for some t 0 > 0. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34 The single most imortant roerty of the MGF is that is uniquely determines the distribution of a random vector: Theorem Assume M X (t) and M Y (t) are the MGFs of the random vectors X and Y and such that M X (t) M Y (t) for all t ( t 0,t 0 ) n.then F X (z) F Y (z) for all z R n where F X and F Y are the joint cdfs of X and Y. Remarks: F X (z) F Y (z) for all z R n clearly imlies M X (t) M Y (t). Thus M X (t) M Y (t) () F X (z) F Y (z) Most often we will use the theorem for random variables instead of random vectors. In this case, M X (t) M Y (t) for all t ( t 0,t 0 ) imlies F X (z) F Y (z) for all z R. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 3/ 34 Connection with moments Let k,...,k n be nonnegative integers and k k + + k n. Then @ k M X (t) E e tx+ +tnxn @t k @tkn n @t k @tkn n @ k E e tx+ +tnxn @t k @tkn n E X k Xkn n e tx+ +tnxn Setting t 0 (0,...,0) T, we get @ k @t k @tkn n @ k M X (t) t0 E X k Xkn n For a (scalar) random variable X we obtain the kth moment of X: d k dt k M X(t) t0 E X k STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 4/ 34

Theorem Assume X,...,X m are indeendent random vectors in R n and let X X + + X m.then Proof: my M X (t) M Xi (t) i M X (t) E e tt X E e tt (X + +X m) E e tt X e tt X m E e tt X E e tt X m M X (t) M Xm (t) Examle: MGF for X Gamma(r, ) and X + + X m where the X i are indeendent and X i Gamma(r i, ). Examle: MGF for X Poisson( ) and X + + X m where the X i are indeendent and X i Gamma( i ). Also, use the MGF to find E(X), E(X ),andvar(x). Note: This theorem gives us a owerful tool for determining the distribution of the sum of indeendent random variables. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 5/ 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 6/ 34 Theorem 3 Assume X is a random vector in R n, A is an m n real matrix and b R m. Then the MGF of Y AX + b is given at t R m by Proof: M Y (t) e tt b M X (A T t) M Y (t) E e tt Y E e tt (AX+b) e tt b E e tt AX e tt b E e (AT t) T X e tt b M X (A T t) Alications to Normal Distribution Let X N(0, ). Then M X (t) E(e tx ) Z e t / e t / Z Z e tx e x / dx e (x tx) dx e (x t) {z } N(t,) df dx Z e (x t) t dx Note: In the scalar case Y ax + b we obtain M Y (t) e tb M X (at) We obtain that for all t R M X (t) e t / STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 7/ 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 8/ 34

Moments of a standard normal random variable Recall the ower series exansion e z P k0 zk valid for all z R. Aly this to z tx X M X (t) E(e tx (tx) k )E k0 X E (tx) k X E X k t k and to z t / Matching the coe k0 M X (t) e t / X i0 k0 t / i cient of t k,fork,,... we obtain 8 >< E(X k ) k/ (k/)!, k even, >: 0, k odd. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 9/ 34 i! Sum of indeendent normal random variables Recall that if X N(0, ) and Y X + µ, where > 0, then Y N(µ, ). Thus the MGF of a N(µ, ) random variable is M Y (t) e tµ M X ( t) e tµ e ( t) / e tµ+t / Let X,...,X m be indeendent r.v. s with X i N(µ i, i ) and set X X + + X m. Then my my M X (t) M Xi (t) e tµi+t i / e t P m i µi + P m i This imlies i i X m X N µ i, i mx i i i.e., X + + X m is normal with mean µ X P m i µ i and variance X P m i i. i t / STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 0 / 34 Multivariate Normal Distributions Linear Algebra Review Recall that an n n real matrix C is called nonnegative definite if it is symmetric and x T Cx 0 for all x R n and ositive definite if it is symmetric and x T Cx > 0 for all x R n such that x 6 0 Let A be an arbitrary n n real matrix. Then C A T A is nonnegative definite. If A is nonsingular (invertible), then C is ositive definite. Proof: A T A is symmetric since (A T A) T (A T )(A T ) T A T A. Thus it is nonnegative definite since x T (A T A)x x T A T Ax (Ax) T (Ax) kaxk 0 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34 For any nonnegative definite n n matrix C the following hold: () C has n nonnegative eigenvalues,..., n (counting multilicities) and corresonding n orthogonal unit-length eigenvectors b,...,b n : Cb i i b i, i,...,n where b T i b i, i,...,n and b T i b j 0if i 6 j. () (Sectral decomosition) C can be written as C BDB T where D diag(,..., n) is the diagonal matrix of the eigenvalues of C, andb is the orthogonal matrix whose ith column is b i,i.e.,b [b...b n ]. (3) C is ositive definite () C is nonsingular () all the eigenvalues i are ositive STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34

(4) C has a unique nonnegative definite square root C /,i.e.,there exists a unique nonnegative definite A such that C AA Proof: We only rove the existence of A. Let D / diag( / /,..., n ) and note that D / D / D. Let A BD / B T. Then A is nonnegative definite and Remarks: A AA (BD / B T )(BD / B T ) BD / B T BD / B T BD / D / B T C If C is ositive definite, then so is A. If we don t require that A be nonnegative definite, then in general there are infinitely many solutions A for AA T C. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 3 / 34 Lemma 4 If is the covariance matrix of some random vector X (X,...,X n ) T, then it is nonnegative definite. Proof: We know that Cov(X) is symmetric. Let b R n be arbitrary. Then b T b b T Cov(X)b Cov(b T X) Var(b T X) 0 so is nonnegative definite Remark: It can be shown that an n n matrix is nonnegative definite if and only if there exists a random vector X (X,...,X n ) T such that Cov(X). STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 4 / 34 Defining the Multivariate Normal Distribution Let Z,...,Z n be indeendent r.v. s with Z i N(0, ). The multivariate MGF of Z (Z,...,Z n ) T is M Z (t) E e ttz E e P n i tizi ny i ny E e tizi i e t i / e P n i t i / e tt t Now let µ R n and A an n n real matrix. Then the MGF of X AZ + µ is M X (t) e tt µ M Z (A T t) e tt µ e (AT t) T (A T t) e tt µ e tt AA T t e tt µ+ tt t where AA T. Note that is nonnegative definite. Definition Let µ R n and let be an n n nonnegative definite matrix. A random vector X (X,...,X n ) is said to have a multivariate normal distribution with arameters µ and if its multivariate MGF is Notation: X N(µ, ). Remarks: M X (t) e tt µ+ tt t If Z (Z,...,Z n ) T with Z i N(0, ), i,...,n,then Z N(0, I), wherei is the n n identity matrix. We saw that if Z N(0, I), thenx AZ + µ N(µ, ), where AA T. One can show the following: X N(µ, ) if and only if X AZ + µ for a random n-vector Z N(0, I) and some n n matrix A with AA T. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 5 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 6 / 34

Mean and covariance for multivariate normal distribution Consider first Z N(0, I), i.e.,z (Z,...,Z n ) T,wheretheZ i are indeendent N(0, ) random variables. Then If X N(µ, ), thenx AZ + µ for a random n-vector Z N(0, I) and some n n matrix A with AA T. and Thus E(Z) E(Z ),...,E(Z n ) T (0,, 0) T 8 < if i j, E (Z i E(Z i ))(Z j E(Z j )) E(Z i Z j ) : 0 if i 6 j We have E(AZ + µ) AE(Z)+µ µ Also, Cov(AZ + µ) Cov(AZ) A Cov(Z)A T AA T Thus E(X) µ, Cov(X) E(Z) 0, Cov(Z) I STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 7 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 8 / 34 Joint df for multivariate normal distribution Lemma 5 If a random vector X (X,...,X n ) T has covariance matrix that is not of full rank (i.e., singular), then X does not have a joint df. Theorem 6 If X (X,...,X n ) T N(µ, ), where is nonsingular, then it has a joint df given by f X (x) ( )n det e (x µ)t (x µ), x R n Proof sketch: If is singular, then there exists b R n such that b 6 0 and b 0. Consider the random variable b T X P n i b ix i : Var(b T X)Cov(b T X) b T Cov(X)b b T b 0 Therefore P (b T X c) for some constant c. If X had a joint df f(x), thenforb {x : b T x c} we should have Z Z P (b T X c) P (X B) f(x,...,x n ) dx dx n But this is imossible since B is an (n )-dimensional hyerlane whose n-dimensional volume is zero, so the integral must be zero. B Proof: We know that X AZ + µ where Z (Z,...,Z n ) T N(0, I) and A is an n n matrix such that AA T. Since is nonsingular, A must be nonsingular with inverse A. Thus the maing h(z) Az + µ is invertible with inverse g(x) A (x J g (x) deta By the multivariate transformation theorem µ) whose Jacobian is f X (x) f Z (g(x)) J g (x) f Z A (x µ) det A STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 9 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 0 / 34

Proof cont d: Since Z (Z,...,Z n ) T,wheretheZ i are indeendent N(0, ) random variables, we have f Z (z) ny e z i / e P n i z i e zt z ( ) n ( ) n i Secial case: bivariate normal For n we have µ µ µ and so we get where µ i E(X i ), i Var(X i ), i,, and f X (x) f Z A (x µ) det A e (A (x µ)) T (A (x µ)) det A ( ) n since det A ( ) n e (x µ)t (A ) T A (x µ) det A ( )n det e (x µ)t (x µ) det and (A ) T A (exercise!) (X,X ) Cov(X,X ) Thus the bivariate normal distribution is determined by five scalar arameters µ, µ,,,and. is ositive definite () is invertible () det > 0: det ( ) > 0 () < and > 0 so a bivariate normal random variable (X,X ) has a df if and only if the comonents X and X have ositive variances and <. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution / 34 We have det Thus the joint df of (X,X ) T N(µ, ) is f(x,x ) e ( ) (x µ ) + (x µ ) (x µ )(x µ ) and (x µ) T (x µ) i x µ hx µ, x µ ( ) x µ h i ( ) x µ, x µ (x µ ) (x µ ) (x µ ) (x µ ) ( ) (x µ ) ( ) (x µ ) (x µ )(x µ )+ (x µ ) + (x µ ) (x µ )(x µ ) STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 3 / 34 Remark: If 0,then f(x,x ) e e (x µ ) (x µ ) f X (x )f X (x ) + (x µ ) e (x µ ) Therefore X and X are indeendent. It is also easy to see that f(x,x )f X (x )f X (x ) for all x and x imlies 0. Thus we obtain Two jointly normal random variables X and X are indeendent if and only if they are uncorrelated. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 4 / 34

In general, the following imortant facts can be roved using the multivariate MGF: (i) If X (X,...,X n ) T N(µ, ), thenx,x,...x n are indeendent if and only if they are uncorrelated, i.e., Cov(X i,x j )0if i 6 j, i.e., is a diagonal matrix. (ii) Assume X (X,...,X n ) T N(µ, ) and let X (X,...,X k ) T, X (X k+,...,x n ) T Then X and X are indeendent if and only if Cov(X, X )0 k (n k),thek (n k) matrix of zeros, i.e., can be artitioned as 0 k (n k) 0 (n k) k where Cov(X ) and Cov(X ). Marginals of multivariate normal distributions Let X (X,...,X n ) T N(µ, ). If A is an m n matrix and b R m,then Y AX + b is a random m-vector. Its MGF at t R m is M Y (t) e tt b M X (A T t) Since M X ( )e T µ+ T for all R n, we obtain M Y (t) e tt b e (AT t) T µ+ (AT t) T (A T t) e tt (b+aµ)+ tt A A T t This means that Y N(b + Aµ, A A T ),i.e.,y is multivariate normal with mean b + Aµ and covariance A A T. Examle: Let a,...,a n R and determine the distribution of Y a X + + a n X n. STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 5 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 6 / 34 For some ale m<nlet {i,...,i m } {,...,n} such that i <i < <i m. Let e j (0,...,0,, 0,...,0) t be the jth unit vector in R n and define the m n matrix A by 3 Then e T i A 6 7 4. 5 e T i e T i m 3 e T i m X X n 3 AX 6 7 6 7 4. 5 4. 5 6 4 Thus (X i,...,x im ) T N(Aµ, A A T ). X i. X im 3 7 5 Note the following: Aµ 6 4 and the (j, k)th entry of A A T is µ i. µ im 3 7 5 (A A T ) jk A (i k th column of ) j ( ) iji k Cov(X ij,x ik ) Thus if X (X,...,X n ) T N(µ, ), then(x i,...,x im ) T is multivariate normal whose mean and covariance are obtained by icking out the corresonding elements of µ and. Secial case: For m we obtain that X i N(µ i, µ i E(X i ) and i Var(X i), foralli,...,n. i ),where STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 7 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 8 / 34

Conditional distributions Let X (X,...,X n ) T N(µ, ) and for ale m<ndefine X (X,...,X m ) T, X (X m+,...,x n ) T We know that X N(µ, ) and X N(µ, ) where µ i E(X i ), ii Cov(X i ), i,. Then µ and can be artitioned as µ µ µ, where ij Cov(X i, X j ), i, j,. Note that is m m, is (n m) (n m), is m (n m), and is (n m) m. Also, T. We assume that is nonsingular and we want to determine the conditional distribution of X given X x. Recall that X AZ + µ for some Z (Z,...,Z n ) T where the Z i are indeendent N(0, ) random variables and A is such that AA T. Let Z (Z,...,Z m ) T and Z (Z m+,...,z n ) T. We want to determine such A in a artitioned form with dimensions corresonding to the artitioning of : B 0 m (n m) A C D We can write AA T as B 0 m (n m) B T C T C D 0 (n m) m D T BB T BC T CB T CC T + DD T STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 9 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 30 / 34 We want to solve for B, C and D. First consider BB T. We choose B to be the unique ositive definite square root of : B / Recall that B is symmetric and it is invertible since is. Then CB T imlies Then CC T + DD T gives C (B T ) B DD T CC T B B ( ) T (BB) Now note that X AZ + µ gives X BZ + µ, X CZ + DZ + µ STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 3 / 34 Since B is invertible, given X x,wehavez B (x µ ). So given X x, we have that the conditional distribution of X and the conditional distribution of are the same. CB (x µ )+DZ + µ But Z is indeendent of X, so given X x, the conditional distribution of CB (x µ )+DZ + µ is the same as its unconditional distribution. We conclude that the conditional distribution of X given X x is multivariate normal with mean E(X X x ) µ + CB (x µ ) µ + B B (x µ ) µ + (x µ ) and covariance matrix DD T STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 3 / 34

Secial case: bivariate normal Suose X (X,X ) T N(µ, ) with µ µ and µ We have and µ + (x µ )µ + (x µ ) ( ) Thus the conditional distribution of X given X x is normal with (conditional) mean and variance E(X X x )µ + (x µ ) Var(X X x ) ( ) Equivalently, the conditional distribution of X given X x is N µ + (x µ ), ( ) If <, then the conditional df exists and is given by f X X (x x ) x µ ( ) e Remark: Note that E(X X x )µ + (x (a ne) function of x. (x µ ) ( ) µ ) is a linear Examle: Recall the MMSE estimate roblem for X N(0, X ) from the observation Y X + Z, wherez N(0, Z ) and X and Z are indeendent. Use the above the find g (y) E[X Y y] and comute the minimum mean square error E (X g (Y )). STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 33 / 34 STAT/MTHE 353: 5 MGF & Multivariate Normal Distribution 34 / 34