Imagine tossing a fair coin 3 times.

What are the probabilities of getting 0, 1, 2 or 3 heads?

H T

H T H T

H T H T H T H T

No. heads 0 1 2 3

Probability 1 3 3 1
8 8 8 8

Many people say that it is just ‘pure chance’, and that you are as likely to get
3 heads in a row as you are to get any other combination, but this is wrong!

This is because there are more combinations that give 1 head or 2 heads
 Consider the number of combinations from tossing coins:

1 coin Outcome 1H 1T
1 pick No. of combinations 1 1 This pattern is known as
Pascal’s Triangle
2 coins Outcome 2H 1H,1T 2T
2 picks No. of combinations 1 2 1

3 coins Outcome 3H 2H,1T 1H,2T 3T

3 picks No. of combinations 1 3 3 1

4 coins Outcome 4H 3H,1T 2H,2T 1H,3T 4T

4 picks No. of combinations 1 4 6 4 1
r 0 r 1 r 2 r 3
5 coins Outcome 5H 4H,1T 3H,2T 2H,3T 1H,4T 5T
5 picks No. of combinations 1 5 10 10 5 1

It gives the number of combinations that Eg how many ways can you
give r matches from n picks between 2 get 3 tails out of 4 tosses?
options (in this case heads or tails) n4 r 3 4 ways
 Pascal realised that the number of combinations This calculation is known as
could be directly calculated from n and r: n
  or n Cr
n! r
Number of combinations 
r!(n  r )! and has a calculator key

Eg how many ways can you Possible combinations:

get 3 heads out of 5 tosses?
HHHTT HHTTH HTTHH TTHHH
5
C3  10 HHTHT HTHTH THTHH
HTHHT THHTH
Eg how many ways can you
THHHT
get 2 heads out of 5 tosses?
This is reflected by the fact that
5
C2  10
Pascal’s triangle is symmetrical
Why are these the same?
1 1
If r = n – r, the calculation gives the same result 1 2 1
ie substitute r = n – r in the rule for nCr 1 3 3 1
1 4 6 4 1
n!
Number of combinations 
 n  r !r! 1 5 10 10 5 1
Where does the rule Cr  r! n  r  ! come from?
n
For example, why does 5 C3 
n! 5!
3!2!
?

ie how many ways can you obtain 3 heads and 2 tails with 5 coins?
Call the 3 heads H1, H2 and H3 and the two tails T1 and T2
Any of the 5 could be the first you toss
Any of the remaining 4 could be the second
Etc until last position Hence 5 x 4 x 3 x 2 x 1 = 5! ways to arrange the 5 coins
But we don’t care which Head came 1st, 2nd or 3rd and which tail came 1st or 2nd

There are 2! ways to order the tails and 3! ways to

Consider the ways order the heads, giving 2! x 3! ‘duplicates’ of HHHTT
to obtain HHHTT:
The same is true for any other ways of
H1H2H3 T1T2 H1H2H3 T2 T1 obtaining 3 heads and 2 tails (eg HHTHT)
H1H3H2T1T2 H1H3H2T2 T1 5
It follows that if C3 = the number of distinct outcomes:
H2H1H3 T1T2 H2H1H3 T2 T1
Duplicates
H2H3H1T1T2 H2H3H1T2 T1
Desired, 5
C3  3!2! 5! 5 C3  3!5!2!
H3H1H2T1T2 H3H1H2T2 T1 distinct
H3H2H1T1T2 H3H2H1T2 T1 outcomes Total number of
possible arrangements
 Probability calculations
Eg a coin is biased so that the probability of tossing heads is 0.3
What is the probability of getting 3 heads out of 5 tosses?

From previously, we know there are C3  10 combinations giving 3 heads

5

The probability of each of those combinations is found by:

3 heads means 2 tails

0.33  0.7 2
P(heads) = 0.3 P(tails) = 1 – P(heads)
Hence the probability of getting 3 heads out of 5 tosses  5C3  0.33  0.7 2  0.1323

In general, the probability of getting r Eg A drug is known to be effective on

60% of patients. What is the probability
out of n outcomes with probability p that 7 out of 10 treatments work?
P (r ) nCr  p r  (1 p ) n  r 10
C7  0.67  0.43  0.2150(4sf)
 Probability distributions
Eg a coin is biased so that the probability of tossing heads is 0.3
What are the probabilities of the different combinations when tossing it 5 times?

P  r   n Cr  p r  1  p 
nr

x P(x)
0 5
C0  0.30  0.7 5  0.1681
1
5
C1  0.31  0.7 4  0.3602
2 5
C2  0.32  0.73  0.3087
3 5
C3  0.33  0.7 2  0.1323
4 5
C4  0.34  0.71  0.0284
5 5
C5  0.35  0.7 0  0.0024

If X is the variable the number of heads obtained, X is said to be Binomially

distributed with probability p of occurring over each of n trials, written X ~ B(n,p)
 Cumulative probability calculations
Eg from previously, X ~ B(5,0.3). What is P(X≤3)?
P( X  3)  P(0)  P(1)  P(2)  P (3)
These calculations would take a long time to do manually, especially if n were larger

The formulae booklet provides tables called the

BINOMIAL CUMULATIVE DISTRIBUTION n  10, p  0.4
FUNCTION by values of n and p, in a similar way
to the Normal Distribution tables you used in S1 x P(X≤x)
0 0.0060
1 0.0464
Eg X ~ B(10,0.4)
2 0.1673
Find: 3 0.3823
a) P(X≤3)  0.3823 4 0.6331
5 0.8338
b) P(X<8)  P( X  7)  0.9877 6 0.9452
7 0.9877
c) P(X≥6) 1 P ( X  5) 1 0.8338  0.1662
8 0.9983
9 0.9999
d) P(X>4)  1 P ( X  4)  1 0.6331  0.3669 Now try Ex1C
 Dealing with p > ½
The tables only give probabilities up to p = ½

They can be adapted to values of p > ½ though

Eg In the production of a car it is found that 85% are without defects.

The cars are produced in batches of 50.
Find the probability that there are at least 40 defect-free cars in a batch.

Define X = number of cars without defects

Then X ~ B(50,0.85) and we require P(X≥40)

There are no tables for p = 0.85, so:

Define Y = number of cars with defects

Then Y ~ B(50,0.15) and we require P(Y≤10) = 0.8801

 Dealing with ‘awkward’ values of p
The tables only give probabilities in increments of 0.05

If p is not a multiple of 0.05, you must calculate manually using:

P  r   n Cr  p r   1  p 
nr

Eg 27% of the A-Level student at MGS study Further Maths.

Find the probability that in a sample of 5 students, at least 3 study Further Maths

Define X = number of students studying Further Maths

Then X ~ B(5,0.27) and we require P(X≥3)

P( X  3)  P(3)  P (4)  P(5)
There are no tables for p = 0.27, P(3) 5C3  0.273  0.732
so calculate manually
P(4)5C4  0.274  0.731 
P(5)5C5  0.275  0.730
 0.1257(4sf)
1. A coin is biased so that P(heads) = 0.8
Find the probability that from 15 tosses, more than 6 are heads
Define X = number of heads in 15 tosses

Then X ~ B(15,0.8) and we require P(X>6)

Define Y = number of tails

Then Y ~ B(15,0.2) and we require P(Y≤8) = 0.9992

2. 1% of the cars manufactured on a production line are defective.

Find the probability that in a sample of 200 cars, more than 2 are defective.

Define X = number of defective cars in sample of 200

Then X ~ B(200,0.01) and we require P(X>2)

P( X  2)  1 P( X  2) P(0)  0.93200
= 0.3233 (4sf) P(1) 200C1  0.071  0.93199 
P(2) 200C2  0.072  0.93198
P( X  2)  0.6766...
 Harder problems
Eg Lesson observations are judged on 9 criteria. A lesson is judged as
outstanding if 5 or more criteria are assessed as outstanding.
Mr Walker achieves outstanding in each criteria with a probability of 0.7
a) Calculate the probability that Mr Walker’s lesson is judged as outstanding overall.

X = number of criteria not outstanding X ~ B 9 , 0.3  P X  4  0.9012

b) Calculate the probability that over the course of 10 observations:
i) none of them are outstanding
X = number of outstanding lessons X ~ B10 , 0.9012
For the Binomial distribution X ~ B n, p  P  X  0  10 C 0  0 .9012 0  1  0 .9012 
10

P  X  r   n C r  p r   1 p 
nr
 0.0000000000 89 (2sf)

P  X  8  10 C8  0.9012 8  1  0.9012 

2
ii) 8 or more of them are
judged as outstanding. P  X  9  10 C9  0.9012 9  1  0.9012  
Now try: P  X  10   0.9012 10
Ex1E, Q5 P  X  8   0.9319 (4dp)
Review Ex1, Q8
 Harder problems
Eg Jan 12, Q3 The probability of a sales representative making a sale on a call is 0.15
Representatives are required to achieve a mean of 5 sales each day.

d) Calculate the least number of calls that need to be made by a

representative for the probability of at least 1 sale to exceed 0.95

If X = number of sales each day

then X ~ B(n,0.15)

P( X  1)  0.95
where P ( X  0)  0.85
n
 P( X  0)  0.05

0.85 n  0.05
 n log 0.85  log 0.05
n log 0.05
log 0.85  18.43... At least 19 calls
 Mean and variance for X~ B(n,p)
An important theme in S2 is the notion of a probability distribution
having measures of average and spread. For a Binomial Distribution:

E  X   np Var  X   np1 p 
Eg X ~ B(20,0.2)  20  0.2  20  0.2  0.8
4  3.2
Eg Jan 12, Q3 The probability of a sales representative making a sale on a call is 0.15
Representatives are required to achieve a mean of 5 sales each day.
c) Find the least number of calls each day a
representative should make to achieve this target.
If X = number of sales each day
then X ~ B(n,0.15) and mean = 0.15n

0.15n  5
 n  0.515  33 31
At least 34 calls
Now try Ex1D, p14
 Poisson Distributions
Many situations fit a different probability model called the Poisson Distribution

The Poisson Distribution is used when events occur : Singly (one at a time)
Independently of each other
Eg the number of accidents Eg the number of hits
At a constant rate
on a certain road each year a website receives in
a 1 hour period
A Poisson Distribution is defined by its mean λ, and is denoted by X ~ Po(λ)

 x
In which case, P  X  x   e .
 The derivation of this rule
x! is beyond the scope of S2
Eg X ~ Po(4). Find:
c) P(X≤1)  P (0)  P (1)
4
a) P(X=0)  e . 40!
0 4
b) P(X=1)  e . 41!
1

 0.0183  0.0733 
 e 4 . 40!  41!
0 1

 0.0916
The formulae booklet provides POISSON CUMULATIVE
DISTRIBUTION FUNCTION tables for λ up to 10 in increments of 0.5

Eg X ~ Po(7). Find: a) P(X≤5)  0.3007 b) P(X>12)  1 P ( X  12)  0.0270

 Poisson approximation of Binomial
If E(X)  Var(X) then B(n,p) can be approximated by Po(np)

This can be used when n is larger than the tables provide

Eg use a Poisson approximation to find P(X≤6) for X ~ B(500,0.02)

B 500 , 0.02  Po10  P X  6   0.1301

When does this give a good approximation?

One of the properties of the Poisson Distribution is that E(X) = Var(X) = λ

You have seen that the Binomial Distribution B(n,p) has:

E  X   np Var  X   np1 p 

So E(X) is close to Var(X) when p is very small, so that 1-p is almost 1.

Conversely, if p is close to poisson
1, the approx.agg
approximation is also better.
The approximation is also better when n is large, for reasons beyond the scope of S2

Now try Review Exercise 1, p64-67, Q2, 6, 13

2a) B 200 , 0.02  Po 4 
x 45
P X  x   e .

P X  5  e .4
 0.1563 (4dp)
x! 5!

b) P  X  5   P X  4   0.6288

6) B150 , 0.02  Po 3

P X  7   1 P X  7   0.0119

13a) n large, p small d) X ~ B1000 , 0.01

b) 0.01  0.0001 E  X   np  10
2

c) X ~ B 5 , 0.01 Var  X   np1 p 

P X  1  1  P X  0   P X  1   10  0.99  9.9

 1 0.995  5C1  0.011  0.99 4  e) B1000 , 0.01  Po10

 0.00098 P  X  6   1 P  X  6 
 0.8699
WB1. In a town, 30% of residents listen to the local radio station. Four residents
are chosen at random.
(a) State the distribution of the random variable X, the number of these four
residents that listen to local radio.

X ~ B 4 , 0.3  (b) On graph paper, draw the

probability distribution of X.
For the Binomial distribution X ~ B n, p 

P  X  r   nCr  p r   1 p 
nr

x P(X=x)
0 0.7 4  0.2401
1
4
C1  0.3  0.73  0.4116
2 4
C2  0.3 2  0.72  0.2646
3 4
C3  0.33  0.7  0.0756
4 0.3 4  0.0081

(c) Write down the most likely (d) Find E(X) and Var (X).
number of these four residents that
listen to the local radio station. E  X   np  4  0.3  1.2

1 Var  X   np1 p   4  0.3  0.7  0.84

WB2(a) State two conditions under which a random
variable can be modelled by a binomial distribution.

2 possible outcomes Probability of success is constant Fixed number of trials

In the production of a certain electronic component it is found that 10% are defective.
The component is produced in batches of 20.
(b) Write down a suitable model for the distribution of defective components in a batch.
X ~ B 20 , 0.1 A supplier buys 100 components.
The supplier will receive a refund
Find the probability that a batch contains if there are more than 15 defective
(c) no defective components, components.
P X  0   0.9 20  0.1216 (4dp) (f) Using a suitable approximation,
find the probability that the
(d) more than 6 defective components. supplier will receive a refund.

P X  6   1 P  X  6   1 0.9976  0.0024 If E  X   Var  X 

(e) Find the mean and the variance of use the approximation
the defective components in a batch. B n, p   Po np 
E  X   np  20  0.1  2 B100 , 0.1  Po10
Var  X   np1 p   20  0.1 0.9  1.8 P X  15   1 P X  15
 1 0.9513  0.0487
WB3 Accidents on a particular stretch of motorway occur at an average rate of 1.5
per week (a) Write down a suitable model to represent the number of accidents per
week on this stretch of motorway.

X ~ Po1.5 

Find the probability that

(b) there will be 2 accidents in the same week,

For the Poisson distribution X ~ Po   2

1 . 5
 x P  X  2   e 1.5 .  0.2510 (4dp)
P X  x   e  . 2!
x!
(c) there is at least one accident per week for 3 consecutive weeks,
P X  1  1 P  X  0   1 0.2231  0.7769
P at least 1 accident per week for 3 weeks   0.77693  0.4689 (4dp)
(d) there are more than 4 accidents in a 2 week period.
Now try Ex2B
Redefine X = number of accidents in 2 week period
X ~ Po 3  P X  4   1 P  X  4   1 0.8153  0.1847
WB4(a) Write down two conditions needed to be able to
approximate the binomial distribution by the Poisson distribution.
If E  X   Var  X 
use the approximation This happens when n is large and p is small
B n, p   Po np 

A researcher has suggested that 1 in 150 people is likely to catch a particular virus.
Assuming that a person catching the virus is independent of any other person
catching it,
(b) find the probability that in a random sample of 12 people, exactly 2 of them
catch the virus.
X ~ B12 , 150
1

P  X  12  12 C 2   150    149 
1 2 10

For the Binomial distribution X ~ B n, p  150

 0.0027 (4dp)
P X  x   nC x  p x  1 p 
n x

(c) Estimate the probability that in a random sample

of 1200 people fewer than 7 catch the virus.

X ~ B1200 , 150
1
  Po 8 P X  7   P X  6  0.3134
Do weaker players have a better chance of winning shorter or longer matches?

For example, does a weaker table-tennis player have a

better chance of winning over three games or five games?

Suppose the weaker player has the same probability p of winning any one game

Let P(x) = the probability of winning in x games

Best of 3 P 2 1C1 pp  p 2
P 3   2C1 p1 p  p  2 p 2 1 p 
P win   p 2 1 21 p    p 2  3  2 p 

Best of 5 P  3   2C 2 p 2 p  p 3
P  4   3 C2 p 2  1  p  p  3 p 3  1  p 
P 5  4C2 p 2 1 p  p  6 p 3 1 p 
2

P win   p 3 1 31 p   61 p    p 3 10  9 p 