Lec 9

MA1061Probability
Lecture 9
Sibylle Schroll
University of Leicester
Sibylle Schroll (University of Leicester)
MA1061Lecture 9
1 / 25
1 / 25
Plan
In the last lecture we talked about

Random variables;
MA1061Lecture 9
2 / 25
2 / 25
Plan

Random variables;
Distribution of random variables.
MA1061Lecture 9
2 / 25
2 / 25
Plan

Random variables;
Distribution functions;
MA1061Lecture 9
2 / 25
2 / 25
Plan

Random variables;
Distribution functions;
A few examples of random variables.
MA1061Lecture 9
2 / 25
2 / 25
Probability mass and distribution functions
Distribution of a random variable

Recall that
A random variable X is a function X : R.
MA1061Lecture 9
3 / 25
3 / 25

Recall that
The probability distribution of X is a new probability PX defined
on the range (RX , R) such that for B R
P(X B) = PX (B) = P(X1 (B)) = P({ : X() B}).
MA1061Lecture 9
3 / 25
3 / 25

Recall that
The probability distribution of X is a new probability PX defined
on the range (RX , R) such that for B R
P(X B) = PX (B) = P(X1 (B)) = P({ : X() B}).
The probability structure of a finite random variable X is
completely specified by its probability mass function
PX () : RX [0, 1],
X
PX (x) = 1.
xRX
MA1061Lecture 9
3 / 25
3 / 25
Distribution function
Also recall cumulative distribution functions:
MA1061Lecture 9
4 / 25
4 / 25
Distribution function
Also recall cumulative distribution functions:

Definition (Distribution function)
Let x R. The function
FX (x) = P{ : X() 6 x},
is called the (cumulative) distribution function of the random variable
X.
MA1061Lecture 9
4 / 25
4 / 25
Basic properties of distribution functions
And some properties:

1
FX () = 0;
MA1061Lecture 9
5 / 25
5 / 25

1
FX () = 0;
2
FX (+) = 1;
MA1061Lecture 9
5 / 25
5 / 25

1
FX () = 0;
2
FX (+) = 1;
3
FX (x) is right-continuous; ie limh0 F(x + h) = F(x).
MA1061Lecture 9
5 / 25
5 / 25

1
FX () = 0;
2
FX (+) = 1;
3
4
FX is non-decreasing.
MA1061Lecture 9
5 / 25
5 / 25

1
FX () = 0;
2
FX (+) = 1;
3
4
FX is non-decreasing.
5
FX is piecewise constant.
MA1061Lecture 9
5 / 25
5 / 25
Comment on random variables
Random variables arise in many different settings;
MA1061Lecture 9
6 / 25
6 / 25

Very different experiments can lead naturally to essentially the
same random variable;
MA1061Lecture 9
6 / 25
6 / 25

The particular random variablesdistributions we shall learn
about in the next few lectures appear in a wide variety of
settings.
MA1061Lecture 9
6 / 25
6 / 25

The particular random variablesdistributions we shall learn
about in the next few lectures appear in a wide variety of
settings.
We shall see certain patterns arising.
MA1061Lecture 9
6 / 25
6 / 25
Particular random variables
Example
Example
In human populations, the probability of male and female are
approximately equal:
1
P(male) = P(female) = .
2
MA1061Lecture 9
7 / 25
7 / 25
Example
Example
1
2
Consider now a family with three kids;
MA1061Lecture 9
7 / 25
7 / 25
Example
Example
1
2
Let X count the number of male children.
MA1061Lecture 9
7 / 25
7 / 25
Example
Example
1
2
Let X count the number of male children.
What is the distribution of X?
MA1061Lecture 9
7 / 25
7 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8
Compare to the distribution of the number of heads when you

toss three coins.
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8

toss three coins.
Different settings but same distribution.
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8

toss three coins.
In both experiments independent trials of a random variable with
two possible values are performed.eg
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8

toss three coins.
Girl or Boy?
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8

toss three coins.
Girl or Boy?
Heads or Tails?
MA1061Lecture 9
8 / 25
8 / 25
Example continued
x
PX (x)
1
8
3
8
3
8
1
8

toss three coins.
Girl or Boy?
Heads or Tails?
0 or 1? Success or Failure?
MA1061Lecture 9
8 / 25
8 / 25
More comments
Different experiments lead to the same random variables;
MA1061Lecture 9
9 / 25
9 / 25
More comments

natural to study a random variable without considering the
underlying experiment;
MA1061Lecture 9
9 / 25
9 / 25
More comments

We may study random variables without specifying the
probability space (, A, P).
MA1061Lecture 9
9 / 25
9 / 25
More comments

Given a random variable X we are only interested in the
probability space (RX , R, PX ).
MA1061Lecture 9
9 / 25
9 / 25
More comments

Given a random variable X we are only interested in the
probability space (RX , R, PX ).
We are interested only in the probability distribution or
distribution function.
MA1061Lecture 9
9 / 25
9 / 25
Discrete uniform distribution equally likely
In many of the examples we considered the n possible outcomes

are equally likely.
MA1061Lecture 9
10 / 25
10 / 25

are equally likely.
Definition (Discrete Uniform Distribution)
Let X take each of the values {x1 , . . . , xn } with the same probability
1
.
n
Then we say that X is uniformly distributed on the set {x1 , . . . , xn }.
The distribution of X is called the discrete uniform distribution.
MA1061Lecture 9
10 / 25
10 / 25

are equally likely.
Definition (Discrete Uniform Distribution)
Let X take each of the values {x1 , . . . , xn } with the same probability
1
.
n
Then we say that X is uniformly distributed on the set {x1 , . . . , xn }.
The distribution of X is called the discrete uniform distribution.
e.g. the distribution of the number showing on a fair die.
MA1061Lecture 9
10 / 25
10 / 25
Bernoulli random variable/Bernoulli distribution
Definition (Bernoulli random variable)

A random variable X that takes the values 1 and 0 with probabilities
p, and q = 1 p resp. is called a Bernoulli random variable.
MA1061Lecture 9
11 / 25
11 / 25

Any experiment with two possible outcomes can be modelled by
a Bernoulli variable. eg the sex of a child, the outcome of a toss
of a coin.
MA1061Lecture 9
11 / 25
11 / 25

Any experiment with two possible outcomes can be modelled by
a Bernoulli variable. eg the sex of a child, the outcome of a toss
of a coin.
How is a Bernoulli random variable related to an indicator?
MA1061Lecture 9
11 / 25
11 / 25
Bernoulli variables as indicators
Let A , in some probability space, such that P(A) = p.
MA1061Lecture 9
12 / 25
12 / 25

Consider X = 1A .
MA1061Lecture 9
12 / 25
12 / 25

Consider X = 1A .
Then P(X = 1) = p, and P(X = 0) = 1 p.
MA1061Lecture 9
12 / 25
12 / 25

Consider X = 1A .
Then P(X = 1) = p, and P(X = 0) = 1 p.
Hence indicators are equivalent to Bernoulli variables.
MA1061Lecture 9
12 / 25
12 / 25
Bernoulli trials and binomial random

variable/binomial distribution
Definition (Sequence of trials)

A sequence of trials or experiments for which
MA1061Lecture 9
13 / 25
13 / 25


(i) the outcomes of the trials are independent;
MA1061Lecture 9
13 / 25
13 / 25


(ii) outcomes are of two types (success, failure); and
MA1061Lecture 9
13 / 25
13 / 25


(iii) the probabilities for the two types of outcomes remain the same
for all the trials,
MA1061Lecture 9
13 / 25
13 / 25


(iii) the probabilities for the two types of outcomes remain the same
for all the trials,
is called an sequence of Bernoulli trials.
MA1061Lecture 9
13 / 25
13 / 25
Sequence of random variables

A sequence of trials is essentially a sequence of independent
random variables X1 , X2 , . . . , Xn .
MA1061Lecture 9
14 / 25
14 / 25

Each random variable must have the same distribution.
MA1061Lecture 9
14 / 25
14 / 25

In the case of Bernoulli trials:
P(X1 = 1) = = P(Xn = 1) = p,
P(X1 = 0) = = P(Xn = 0) = q = 1 p.
MA1061Lecture 9
14 / 25
14 / 25

In the case of Bernoulli trials:
P(X1 = 1) = = P(Xn = 1) = p,
P(X1 = 0) = = P(Xn = 0) = q = 1 p.
Plenty of examples. BUT we still need to define what
independence means in this situation.
MA1061Lecture 9
14 / 25
14 / 25
Independent Bernoulli variables
Suppose now that the events A1 , . . . , An are mutually

independent, and P(Ai ) = p.
MA1061Lecture 9
15 / 25
15 / 25

Can you think of a way to define independent Bernoulli
variables?
MA1061Lecture 9
15 / 25
15 / 25

Can you think of a way to define independent Bernoulli
variables?
Xi = 1Ai , for i = 1, . . . , n.
MA1061Lecture 9
15 / 25
15 / 25
When we have n Bernoulli trials the sample space is

= { : = (a1 , . . . , an ), ai = 0 or 1}.
MA1061Lecture 9
16 / 25
16 / 25

= { : = (a1 , . . . , an ), ai = 0 or 1}.
By independence of trials the probability of an elementary event
in which there are precisely k successes and n k
failures is just
k nk
P() = P((a1 , . . . , an )) = p q
k=
n
X
ai .
i=1
MA1061Lecture 9
16 / 25
16 / 25

= { : = (a1 , . . . , an ), ai = 0 or 1}.
failures is just
k nk
P() = P((a1 , . . . , an )) = p q
k=
n
X
ai .
i=1
How many of these are there?
MA1061Lecture 9
16 / 25
16 / 25

= { : = (a1 , . . . , an ), ai = 0 or 1}.
failures is just
k nk
P() = P((a1 , . . . , an )) = p q
k=
n
X
ai .
i=1
How many of these are there?

Number of orderings of k 1s and (n k) 0s
Cn
k
MA1061Lecture 9
16 / 25
16 / 25
Towards the binomial distribution

On the probability space we just constructed define the variable
X : R by
MA1061Lecture 9
17 / 25
17 / 25

X : R by
X() = X((a1 , . . . , an )) =
n
X
ai = number of successes.
i=1
MA1061Lecture 9
17 / 25
17 / 25

X : R by
X() = X((a1 , . . . , an )) =
n
X
i=1
Thus if in the elementary event there are exactly k successes,

then X() = k.
MA1061Lecture 9
17 / 25
17 / 25

X : R by
X() = X((a1 , . . . , an )) =
n
X
i=1

then X() = k.
We want to find the (probability) distribution of X. Range?
.
MA1061Lecture 9
17 / 25
17 / 25

X : R by
X() = X((a1 , . . . , an )) =
n
X
i=1

then X() = k.
{0, 1, . . . , n}.
MA1061Lecture 9
17 / 25
17 / 25

X : R by
X() = X((a1 , . . . , an )) =
n
X
i=1

then X() = k.
{0, 1, . . . , n}.
Let k = 0, . . . , n. What is PX (k)?
MA1061Lecture 9
17 / 25
17 / 25
By definition

PX (k) = P { : X() = k} .
MA1061Lecture 9
18 / 25
18 / 25
By definition

PX (k) = P { : X() = k} .
Let Bk = { : X() = k}.
MA1061Lecture 9
18 / 25
18 / 25
By definition

PX (k) = P { : X() = k} .
Let Bk = { : X() = k}.
Obviously Bk consists of the binary sequences of length n with
exactly k 1s and n k 0s.
MA1061Lecture 9
18 / 25
18 / 25
By definition

PX (k) = P { : X() = k} .
Let Bk = { : X() = k}.
Obviously Bk consists of the binary sequences of length n with
exactly k 1s and n k 0s.
We just saw that there are Cn
k of these.
MA1061Lecture 9
18 / 25
18 / 25

If then P() = P((a1 , . . . , an )) = pk qnk .
MA1061Lecture 9
19 / 25
19 / 25

Therefore
k nk
.
P({X = k}) = PX (k) = Cn
kp q
MA1061Lecture 9
19 / 25
19 / 25

Therefore
k nk
.
P({X = k}) = PX (k) = Cn
kp q
The variable X is completely characterized by its probability
distribution
k nk
PX (k) = Cn
,
kp q
MA1061Lecture 9
k = 0, . . . , n.
19 / 25
19 / 25

Therefore
k nk
.
P({X = k}) = PX (k) = Cn
kp q
distribution
k nk
PX (k) = Cn
,
kp q
k = 0, . . . , n.
The above distribution is known as the binomial distribution.
MA1061Lecture 9
19 / 25
19 / 25

Therefore
k nk
.
P({X = k}) = PX (k) = Cn
kp q
distribution
k nk
PX (k) = Cn
,
kp q
k = 0, . . . , n.
The above distribution is known as the binomial distribution.

We write X Bi(n, p).
MA1061Lecture 9
19 / 25
19 / 25
Binomial random variable
Definition (Binomial random variable)

A random variable X such that
k
nk
P(X = k) = Cn
,
k p (1 p)
for some p [0, 1], and k = 0, 1, . . . , n, is called a Binomial random

variable with parameters n, p.
The distribution of X defined as above is called the Binomial
distribution with parameters n, p.
MA1061Lecture 9
20 / 25
20 / 25
Binomial distribution and number of successes
MA1061Lecture 9
21 / 25
21 / 25
If X Bi(n, p), then X counts the number of successes in n

independent experiments each with probability of success p.
MA1061Lecture 9
21 / 25
21 / 25
If X Bi(n, p), then X counts the number of successes in n

independent experiments each with probability of success p.
Its distribution PX (k) gives the probability of having k successes
in n independent trials.
MA1061Lecture 9
21 / 25
21 / 25
Theorem
Let X1 , X2 , ..., Xn be n independent Bernoulli random variables of
parameter p. Then
X = X1 + X2 + + Xn
has a binomial distribution. In fact
X Bi(n, p.)
MA1061Lecture 9
22 / 25
22 / 25
Example
Suppose a family has three children and suppose that it is equally
likely to have boy or girl.
If X is the number of male children, what is the probability
distribution of X?
Using this distribution, calculate the probability that the family has 0,
1, 2 or 3 boys.
MA1061Lecture 9
23 / 25
23 / 25
Example
Suppose 10 % of items produced in a factory are defective. Choose 3
items at random. What is the probability that
MA1061Lecture 9
24 / 25
24 / 25
Example
(a) all are defective
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
(c) none are defective?
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
Use the binomial distribution to answer the question.
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
Could you answer the question without using the binomial
distribution?
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
distribution?
X Bi(3, 0.1)
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
distribution?
X Bi(3, 0.1)
Hence P(X = 3) = C33 (0.1)3 (0.9)0 = 0.001
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
distribution?
X Bi(3, 0.1)
Hence P(X = 3) = C33 (0.1)3 (0.9)0 = 0.001
P(X = 2) = C32 (0.1)2 (0.9)1 = 3x0.01x0.9 = 0.027
MA1061Lecture 9
24 / 25
24 / 25
Example
(b) 2 are defective
distribution?
X Bi(3, 0.1)
Hence P(X = 3) = C33 (0.1)3 (0.9)0 = 0.001
P(X = 2) = C32 (0.1)2 (0.9)1 = 3x0.01x0.9 = 0.027
P(X = 0) = C30 (0.1)0 (0.9)3 = 0.729
MA1061Lecture 9
24 / 25
24 / 25
Whats next?
random vectors;
MA1061Lecture 9
25 / 25
25 / 25
Whats next?
random vectors;
independence of random variables;
MA1061Lecture 9
25 / 25
25 / 25
Whats next?
random vectors;
independence of random variables;
functions of random variables.
MA1061Lecture 9
25 / 25
25 / 25

Lec 9

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Lec 9

Enviado por

Direitos autorais:

Formatos disponíveis

MA1061Probability

Sibylle Schroll (University of Leicester)

In the last lecture we talked about

Sibylle Schroll (University of Leicester)

In the last lecture we talked about

Sibylle Schroll (University of Leicester)

In the last lecture we talked about

Sibylle Schroll (University of Leicester)

In the last lecture we talked about

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Distribution of a random variable

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Distribution of a random variable

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Distribution of a random variable

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Also recall cumulative distribution functions:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Also recall cumulative distribution functions:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Basic properties of distribution functions

And some properties:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Basic properties of distribution functions

And some properties:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Basic properties of distribution functions

And some properties:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Basic properties of distribution functions

And some properties:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Basic properties of distribution functions

And some properties:

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Comment on random variables

Random variables arise in many different settings;

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Comment on random variables

Random variables arise in many different settings;

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Comment on random variables

Random variables arise in many different settings;

Sibylle Schroll (University of Leicester)

Probability mass and distribution functions

Comment on random variables

Random variables arise in many different settings;

Sibylle Schroll (University of Leicester)

Particular random variables

Sibylle Schroll (University of Leicester)