Scribd
Upload
57 views10 pages

16.584: Lecture 5: Functions of Random Variables Contd.: Y N I 1 F (X) G (X) XY 2 2

This document discusses functions of random variables. It explains that the probability distribution of a function of one or two random variables can be found by determining the region in the input variable space that maps to values of the function below a given threshold. For functions of two independent random variables, the probability density function is the convolution of the individual densities. The Jacobian of the transformation between the input and output variable spaces is used to relate their probability densities.

Uploaded by

Eric Kial
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
57 views10 pages

16.584: Lecture 5: Functions of Random Variables Contd.: Y N I 1 F (X) G (X) XY 2 2

This document discusses functions of random variables. It explains that the probability distribution of a function of one or two random variables can be found by determining the region in the input variable space that maps to values of the function below a given threshold. For functions of two independent random variables, the probability density function is the convolution of the individual densities. The Jacobian of the transformation between the input and output variable spaces is used to relate their probability densities.

Uploaded by

Eric Kial
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

1

16.584: Lecture 5 : Functions of Random Variables Contd. Functions of one RV : Y = g (X ) fY (y ) =


N f X (x i ) i=1 g (x ) i

Functions of two RVs: Z = g (X, Y ) , with fXY (x, y ) specied. ExampleZ = X + Y Z = XY Z = [X 2 + Y 2] Approach: Find the point set Cz in the (x, y ) plane such that: P [Z z ] = [(X, Y ) Cz ] FZ (z ) =
(x,y )Cz

fXY (x, y )dxdy

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

Ex: Z = X + Y
Y z X=Z-Y z X

Figure 1:

CDF: FZ (z ) = =
x+y z

fXY (x, y )dxdy dy

+ z y y = x= fXY (x, y )dx

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

PDF: dFZ (z ) fZ (z ) = dz + = y= fXY (z y, y )dy If X, Y Independent: fXY (x, y ) = fX (x) fY (y ) fZ (z ) = =


+ y = fX (z + x= fY (z

y ) fY (y )dy x) fX (x)dx

Under independence: Z = X + Y : fZ (z ) = fX (x) fY (y ): Convolution

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

Ex: Z = aX + bY a, b > 0 Let Z = aX + bY = V + W where V = aX and W = bY Note : fV (v ) = fZ (z ) =


1 a fX (v/a)

and fW (w) =
v) fY [ (z b ]dv

1 b fY (w/b)

If X, Y independent:
1 + ab v = fX (v/a)

If fX (x) = fY (y ) = 1 0 < x, y < 1, nd fZ (z )

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

Two Functions of two RVs

V = g (X, Y ) W = h(X, Y ) Given fXY (x, y ) nd the joint pdf fV W (v, w) Approach: From the set of equations v = g (x, y ) and w = h(x, y ) generate the inverse mapping: x = (v, w) y = (v, w) Assume that every (x, y ) maps into a unique (v, w) and vice-versa

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

Consider P [v < V v + dv, w < W w + dw]


w w+dw w p1 p2 y p4 v constant p3 v+dv : constant w+dw: constant w constant v v v + dv x

Figure 2:

Coordinates of the Image in (X-Y) plane (v, w) (v + dv, w) (v, w + dw) (v + dv, w + dw) (x = (v, w), y = (v, w)) ((v + dv, w), (v + dv, w) ((v, w + dw), (v, w + dw)) ((v + dv, w + dw), (v + dv, w + dw))

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

p1 = (x, y ) dv ) p2 = (x + dv, y + v v dw, y + dw) p3 = (x + w w dw), y + dv + dw) p4 = (x + dv + v w v w

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

To determine P [v < V v + dv, w < W w + dw] Let A(Ruw ) and A(Rxy ): innitesimal areas identied in the (v, w) and (x, y ) domains respectively P [v < V v + dv, w < W w + dw] = fV W (v, w) A(Ruw ) = A(Rxy ) fXY (, )d d = fXY (x, y )A(Rxy ) A(Rxy ) fV W (v, w) = fXY (x, y ) A(Rvw ) Ratio
A(Rxy ) A(Rvw )

| [Refer App. C in text] = |J

is the Jacobian of the transformation, x = (v, w), y = where J (v, w) = J


v w v w
c Prof. K. Chandra, October 14, 2008 16.584: Probability and Random Processes; ECE, UMASS Lowell

In general, when multiple disjoint regions Ai(Rxy ) result from A(Ruw ), fV W (v, w) = = 1| where |Ji| = |J i = J
g g x y h h y y n i=1

i| fXY (xi, yi) |J

n i=1 fXY (xi , yi )

|Ji|

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

10

Examples:

If X, Y are independent RVs and R = (X 2 + Y 2) = tan1(Y /X ), nd joint distribution fR(r, ) Assume X, Y : N (0, 1) If Z = aX + bY and W = cX + dY , nd fZW (z, w) in terms of fXY (x, y ) If Z = X/Y and X, Y are independent exponentially distributed RVs with = 1, nd fZ (z )

c Prof. K. Chandra, October 14, 2008

16.584: Probability and Random Processes; ECE, UMASS Lowell

You might also like