LECTURE 9: Conditioning on an event; Multiple continuous r.v.

's
• Conditioning a r.v. on an event
- Conditional PDF
- Conditional expectation and the expected value rule
- Exp ,onential PDF: memorylessness
- Total probability and expectation theorems
- Mixed distributions
• Jointly continuous r.v.'s and joint PDFs
From the joints to the marginals
- Uniform joint PDF example

- The expected value rule and linearity of expectations

- The joint CDF
1
 Conditional PDF, given an event

Px(x) = P(X = x) fx(x) · 8 ~ P(x $ X $ x + 8)

PXIA(x) = P(X = x I A) fx1A(x) · 8 ~ P(x <X < x + 8 I A)

P(X E B) = E Px(x) P(X EB)= ltx(x)dx

xEB

P(X EB I A)= E PXIA(x)

xEB
P(X EB I A)= l fx1A(x) dx

j fxiA(x) dx =1

2
Cond-tiona PDF of X, g-ve t at X E A

P(x S XS x+<S IX EA)~ IXIXEA(x) ·8

0,
lx lXEA(x) = fx(x)
i XE A
P(A)'

3
Conditional expectation of X, given an event

E[X] = Lxpx(x) E[X] =/ xfx(x) dx

X

E(X IA]= LxPx1A(x)

X E[X I A] = Jxfx 1A(x) dx
Expected value rule:

E[g(X)] = Eg(x)px(x)
X
E(g(X)] = f g(x)Jx(x) dx

E(g(X) IA] = Lg(x)pX IA(x) E [g(X) IA] = / g(x)fx 1A(x) dx

X

4
 xample
A: a b<X<b
2 -
f x(x),

-
-
a b C d X

EX I A]=

E X 2 I A]=

a b C d X

5
 Memorylessness of the exponential PDF

• Do you prefer a used or a new ..exponential" light bulb? Probabilistica y identical!

• Bulb lifetime T: exponential(,X)

P(T > x) = e-Ax, for x ~ 0

- we are told that T >t
- r.v. X: remaining lifetime
P (X > x IT > t) = e-Ax, for x ~ 0

6
Memorylessness of the exponential PDF

P(O 5-.T 5-.8)

P(t 5-.T 5-.t + 8 I T > t)

similar to an independent coin flip,

every 8 time steps,
with P(success)~ A8

7
 pro ability and expectatio heo ems

P(An) (BI An)

fx(x) - P(A1) f IA (x) · P(An) f A i(x)

,,,___
__ P_
(A_2_
) E X I A2

E[ - (A1) E XI 1 + + (An) E[X I An

8
 Example

• Bill goes to the supermarket shortly, with probability 1/3,

at a time uniformly distributed between O and 2 hours from now .;
or with probability 2/3, later in the day
at a time uniformly distributed between 6 and 8 hours from now

fx(x)

0 1 2 6 7 8

9
 uniform on 0, 2], with proba bi rty 1 /2 Is X discrete?
X=
1, wit .h proba bi ·ty 1/2
Is X co ti uous?

Y d·s.crete Y, with probabi ity p

x- z, with probab·lity 1 - p
Z co tinuous

Fx(x) =

E[X -

10
 uniform on 0, 2], with proba bi ·ty 1 /2
X=
1, with proba bi rty 1/2

Fx(x)

0 1 2 X

0 1 2 X

0 1 2 X
11
Jointly continuous r.v.'s an jo-nt PDFs Px(x) fx(x)

Px,y(x, y) fx,y(x, y)

Px,y(x, y) - P(X - x and Y - y) ~ 0 fx,y(x, y) >0

P((X,Y) EB)- LL Px,y(x,y) ( ( X, Y) E B)- j j f x y ( x, y) dx dy

(x,y)EB x,y)EB

X y
J:1: fx,y(x, y) dx dy - 1

Defin-tion: Two random variables are Joint y cont-nuou

i they C n be described by a JO PDF

12
((X,Y) EB)= jj fxy(x,y)dxdy
X y)EB

13
 ((X, Y) EB) - j J fxy(x,y)dxdy
X y)EB

(a < X < a+~, c <Y <c t5) ~ f x y(a, c) t52

fxy(x,y). probabirty per u it area

'

area(B) =0 => P((X, Y) EB) =0

14
From the joint to the marginals

Px(x) = LPx,y(x,
y
y) fx(x) = Jfx,y(x,y) dy

py(y) = LPx,y(x,
X
y) fy(y) - j f x,y(x, y) dx

15
Uniform joint PDF on a set s fx(x) - j f x,y(x, y) dy

• !
fy(y) - j f x ,y(x, y) dx
I

J •cw •1. •.

..
'I

I·

16
More than two random variables

PX,Y,z(x, Y, z) f X,Y,z(x, Y, z)

Px(x) = LLPx,Y,z(x,y,z)
y z

Px,y(x,y) = LPx,Y,z(x,y,z)
z

17
Functions of m ltip e andom varia es

Z - g(X Y)

Expected va u ru e:

E[g(X, Y)] = __
X y
g(x, Y)Px,y(x, y) E[g(X, Y) =j j g(x, y)fx,y(x, y) dx dy

inea -ty of expectat-ons

E(aX+ b] = aE[X] b

E[X + Y] = E X] E[Y]

· + E[Xn]
18
The joint CDF

Fx(x) = P(X ~ x) = J: x(t)

00
f dt fx ( x ) =- dFx(
dx
x)

Fx,y(x, y) = P(X < x, Y < y)

82Fxy
fx,y(x, y) = 8x8~ (x, y)

19
MIT OpenCourseWare
https://ocw.mit.edu

Resource: Introduction to Probability

John Tsitsiklis and Patrick Jaillet

The following may not correspond to a particular course on MIT OpenCourseWare, but has been provided by the author as an individual learning resource.

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

20