Dee 5 2022 Lecture 22 MAT 344

Example Suppose we toss and let H and T denote

a coin
the outcomes heads and tails Suppose further that the
coin is such that the probability of the outcome H is p
This means
of that the probability of the coin coming
course
tails after the toss is
up as
q tp
Our sample spare is
S H T
random X rule
Define a
X H
variable
L X T O
on S P
by the
Then
E X
p
for
Ex o
q 1 p p

Linearity of enpatation
Lin
probability space SP
a and suppose X and
Y are two random variables on S P Let
S si sd and pi P si it d Then
E Xt y
II X Isi 4 si pi

Iit
Pi X si Eid Pi Y si
E x E Y
show that LE IR is constant
Similarly one can if any
then
E 2x a EX
Putting thesetogether we see that Xu are random
if Xis
variables on S P and an an ER are real constants
then
linearityof E x X t tank a L E Xi t t an E Xu
expectation
 Independence
Let S P be a probability space Two events A and B
are said to be independent if
PCAA B PIA P B
If PCB then A and B
O are independent if
PCA B PIA
A collection said to be
of events As Au are

mutually independent if
P Ai A A Aip
JI P Aig
for any subset in iz in of In

Example Suppose that

tossing a certain coin the probability
on
heads occurring is p so that the probability of tails
of
is 1 p Assume o pct so that p g o
q
None suppose we toss the and coin a second time
make sure that the first toss does not affect the second
toss in
anyway
The sample space with the notation being obvious
for this experiment of two tosses is
S H H HT T H TT
Let A be the event that the first toss is heads and
B the event that the secondtoss is tails Since the
first toss does not affect the outcome of the second and
vice versa
P BIA P B
q
This means
P IAMB PCARB
q PCB
A p IA p
Plan B P A PB
pq
This means that A and B are independent
 In fact all these pairs of emails are independent
1 A my n it B my y H
2 A my n H B my y T This is the aboveexample
3 A T B H
my ly ny g
4 A ny y t B ny y t

One can follows Fors the coin n times

generalise this as

in a manner that no toss's outcome is influenced

by any
the earlier tosses Now the sample space is
of S Rika nul nie it T i b in
i e S is the set H T stirrings
of length n
Let
of
Ei a m Est ni it i ti in
Fi n Nuts I ni F i Is on
Let it ik be k distinct elements in Eli in and
ji ja k the remaining elements of Is in ie
In in air
ji ja r
Then it is to see that
easy
Ei
Erik Fj i s Fjk
are
mutually independent
Independent random variables
Lt X Y be two random variables on a probability
S P We X and Y are independent
space say if
for any two real numbers x and
P X R
y
Y y P X R PI Y y
The symbols X R Y y Xen Y
y are shorthand for
se SI XIs n and 41s y SES XIs n SE SI YIN y

respectively
In general if W is any random variable and D is
SE SI Wis ED
any set in IR then WED
 There is another of viewing this
Let f x 4
way
be the map
f S s R2

given by
fis Xis 4157
Let
T X is X Y S
Then f's CT and T is finite From a discussion
in Lecture 20 we conclude that CT Pt is a probability
space where P D P f D for all subsets D
of T

n o o a a

o o o o o o o o

d o o o o o o o o

É Vertical and horizontal

o o o o o a lines are independent
o o o o a a in CT P
u o o o o a

a XIs s

In otherwords every horizontal lineis independent of

every
vertical line in the probability space CT Pt More precisely
every
set of dots on a horizontal line is independent of every set
vertical line
of dots on a
It is to see that X and Y are independent
easy
and only if
if
PHEU Y EV PRE U PLYEU
U V in R Here CXEU YEV
for every pair of subsets
is as above the event in S
Xe u Y EV SES X is EU and VISEU
CX EU A CY EV
Remark The events Xi X EU etc are quite obviously
Xi X i
Xen X IU
In the above example with f S s the map
fin X's Y's it is
easy to see that
X U Y V f Ux V
and
Chi Y j f i
j
The probability distribution induced randomvariable
by a

Let X S R be a random variable on a probability

space SP Let Ff S Define

p PEX t te T
In other words p P X It AET It is clear that
It Pt 1

Thus we get a probability measure Px on TICS This is

really old friend P
our Px is often called the
probability distribution of X
Note that
E X t Pt
Ey
This is usually most useful when X s C Z the set
of integers
Bernoulli trials
Any experiment in which there are exactly two outcomes
the two outcomes NEED not be equiprobable is called a
Bernoulli trial The example in this lecture ie the
very first
coin tossing example is an example a Bernoulli trial
We
of
usually identify the sample space of a Bernoulli
 trial with the set Gig and set
p PCD q Plo
Note t
q p
What we have described is a Bernoulli trial with parameter p or
with probability ofsuccess p the outcome 1 is often termed
success and 0 failure
Bernoullirandom variable
Let IS P be a probability space and X S s
big a random
variable and
say P CK1 p As discussed earlier this makes
a Bernoulli
O I a probability
space in fact the sample space of
trial with parameter p Such an X is called a Bernoulli
random variable with parameter p or with probability success p
of
It is easy to see that if X is a Bernoulli random
variable ther
E X p
The
completeno Extort's pax D
Otp p

The Bernoulli distribution

Let me IN and pe to I Consider the probability space S P
where S So T
of binary strings oflength n and
the set
P the probability measure
P x na say
given
i by n
pi CI p Nz m E S
where i is the number of successes ie the numberofI's
in the binary sequence Renz Nn
Why does this define a probability measure To prove that we
have to shove that
Pal 1
Eg
We will do this now
 Let ie 0,1 n Let
Si ni ones se mn contains exactly i 1 s
ie Si is the set n with
of binary strings of length
1 successes None
exactly
Isil ni
Hence
i
P Si ni pi i p

Since S Si and Si
nsj 0 if itj we
get
2
SES
Pest
In Egg Pls
i
If Ies pic pin

Gp i
If pi E L
i
If 1 pic pin
pt 2 p Binomial Theorem

Nent consider the sets

ones mes t
Aj a
nj o
Bj a
nj
for j b sn
S and 10 Hence
None Aj 0Bj Aja Bj
I P
PCAj Bj j b gu
symmetry it is clear that
By PCB P Bn
PCB
Let us work ont PCB
If a an E B has i I's in it then na Nn has
 i 1 I's in it These i 1 I's be the n t
can
any of
spots from 2 to n Thus there are II
It followsthat
binarystrings
in B with i I's in them
PCB i
II II pi Ci p
i
p
Iz II pi i p

Make thesubstitution

É
i k
P I pk i pen k i i

p p i p Binomial Thin

It follows that
P
Aj I p P Bj p j b in

Fur is je n let
S s 0,1
Xj
be the random variable
ni
Xj mn
Nj
Then Xj 1
Bj and Xj 0 Aj
Thus
PCXj 1 p and P
Xj 07 1p
In other words Bernoulli random variable with
Xj is a

probability of success p for every j in El on

It is
easy to see that X Xz xn are independent
left as an exercise Andthey are all Bernoulli with
parameter p They are what are called independentidentically
distributed randomvariables i i d for short
 The random variable
X Kit Xz t Xu
counts the number of successes ie
X aka m
of 2 s contained in 2 on
It is clear from our discussion that
i
P X i I pi i
p i o p

X is called a Binomial random variable with parameters

n
p
It induces a probability space
Oslo su Px
such that
i
Px Ii ni pi t p i o m

The probability space 0,1 in Px is called the

Binomial distribution with parameters nsp

Theorem Let X be the Binomial random variable with

parameters n
p Then
EX up
Proof
We have X X t Xu where Xi are Bernoulli with paamatu p
So E X EE E exit IE p np

Remark The probability space SP above of

binary strings
i what
of length n with Pfa m pi t pin i ofsuccesses
can be regarded the spareof outcomes
as
of n repetitious
of a Bernoulli trial with parameter p te g tossing a coin
n times in such a that the outcome of one of the
way
repeated trials does not affect of the later trials any
 Independent random variables again
Let X and Y be independent random variables on
a probability space S P Let
pi PA i i E X S
PM j Y s
q je
Then
E XY Pai Y j
Iggy if

ij Pax i PG
j since X andY are indep
Iggy

if Pi Gj
Iggy
Texas i pi Zeus I 8j
E X E Y

More generally
using the same strategy one shows that
Xi Xi Xu are independent random variables then
if E X Xz Xu E X E X2 E Xn
We record this as a theorem

ma
random variables on a
III Testament
probability space SP Then