Chapter 2

Discrete Random Variables

2.1 Introduction
A random variable, denoted by a capital letter such as X, is a function mapping from
each outcome in a sample space to the real line:

X : S → R.

Random variable is discrete if its range is either finite or countably infinite.

HHH THH 1 HHH THH 3

HHT THT 2 HHT THT 2
HTH TTH HTH TTH
HTT TTT 3 HTT TTT 1
0
random variable, Z random variable, X

{ {
1, if S = HHH 3, if three heads
Z = 2, if S = {HTT, THT, TTH} X = 2, if two heads
3, otherwise 1, if one head
0, if no heads

Figure 2.1: Examples of random variables

Exercise 2.1 (Introduction)

1. Number of heads, X in three coin tosses: (i) discrete (ii) continuous

domain: (i) S ∈ {TTT, TTH, . . . , HHH} (ii) X ∈ {0, 1, 2, 3}
range: (i) S ∈ {TTT, TTH, . . . , HHH} (ii) X ∈ {0, 1, 2, 3}
Random variable X (i) is (ii) not a probability

2. Number of dots, X, in roll of a die: (i) discrete (ii) continuous

domain: (i) S ∈ dots on die faces (ii) X ∈ {1, 2, 3, 4, 5, 6}

31
32 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

range: (i) S ∈ dots on die faces (ii) X ∈ {1, 2, 3, 4, 5, 6}

Value of random variable X denoted (i) X = x (ii) X = X

3. Number of seizures, Z, in a year: (i) discrete (ii) continuous

domain: (i) S ∈ patients in study (ii) Z ∈ {0, 1, 2, . . .}
range: (i) S ∈ patients in study (ii) Z ∈ {0, 1, 2, . . .}

Patients, W , sick (1 or more seizures) or not (0 seizures):

(i) discrete (ii) continuous
domain: (i) S ∈ patients in study (ii) W ∈ {0, 1}
range: (i) S ∈ patients in study (ii) W ∈ {0, 1}

4. Waiting time, T , at Burger King up to 1 hour: (i) discrete (ii) continuous

domain: (i) S ∈ waiting customers (ii) {T : 0 ≤ T ≤ 1}
range: (i) S ∈ waiting customers (ii) {T : 0 ≤ T ≤ 1}

2.2 Probability Mass Functions

If R is the range of discrete random variable X, the probability mass function (pmf )
of X is function f : R → R which satisfies:

• f (x) > 0, for all x ∈ R

P
• x∈R f(x) = 1
P
• If X ⊂ R, then P (X ∈ A) = x∈A f (x)

The pmf can be described by table, graph or function, these descriptions are called
the distribution of the random variable.

HHH THH 1 1/8 HHH THH 3 1/8

HHT THT HHT THT 2 3/8

2 3/8
HTH TTH HTH TTH
HTT TTT 3 4/8 HTT TTT 1 3/8
0 1/8

random variable, Z probability mass function (pmf ), f(z) random variable, X probability mass function (pmf ), f(x)

{ {
3, if three heads 1/8, if x = 3

{ 1, if S = HHH
Z = 2, if S = {HTT, THT, TTH}
3, otherwise {
1/8, if z = 1
f(z) = 3/8, if z = 2
4/8 , if z = 3
X = 2, if two heads
1, if one head
0, if no heads
f(x) = 3/8, if x = 2
3/8 , if x = 1
1/8, if x = 0

Figure 2.2: Examples of random variable and probability mass function (pmf)

The cumulative distribution function (cdf ) (distribution function) of discrete random

Section 2. Probability Mass Functions (LECTURE NOTES 3) 33

variable X is
X
F (b) = P (X ≤ b) = f (x), for all real numbers b.
x≤b

If range R of discrete random variable X has k elements, X has a uniform distribution

with pmf
1
f (x) = , for each x ∈ R.
k

The mode of discrete random variable X is the value of X where the pmf is a maxi-
mum. The median is the smallest number m such that

P (X ≤ m) ≥ 0.5 and P (X ≥ m) ≥ 0.5.

A random variable with two modes is bimodel, with two or more modes, multimodal.

Exercise 2.2 (Probability Mass Functions)

1. Probability mass function for seizures. The number of seizures, X, of a typical

epileptic person in any given year is given by the following pmf.

x 0 2 4 6 8 10
f (x) 0.17 0.21 0.18 0.11 0.16 0.17

(a) The chance a person has 8 epileptic seizures is f (8) =

(i) 0.11 (ii) 0.16 (iii) 0.17 (iv) 0.21.
(b) The chance a person has at most 4 seizures is
(i) 0.17 (ii) 0.21 (iii) 0.56 (iv) 0.67.
(c) P (X ≤ 4) = (i) 0.17 (ii) 0.21 (iii) 0.56 (iv) 0.67.
(d) f (2) = (i) 0.17 (ii) 0.21 (iii) 0.56 (iv) 0.67.
(e) f (2.1) = (i) 0 (ii) 0.21 (iii) 0.56 (iv) 0.67.
(f) P (X > 2.1) = (i) 0.21 (ii) 0.38 (iii) 0.56 (iv) 0.62.
P10
(g) x=0 f (x) = P (X = 0) + P (X = 2) + · · · + P (X = 10) =
(i) 0.97 (ii) 0.98 (iii) 0.99 (iv) 1.
(h) Graph of distribution. Which of the three graphs best describes the prob-
ability distribution of the number of seizures?
(i) (a) (ii) (b) (iii) (c).
(i) Probability mass function. Which one of the following functions describes
the pmf of the number of seizures?
34 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

P(X = x) P(X = x) f(x)

0.20 0.20 0.20

0.15 0.15 0.15

0.10 0.10 0.10

0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
x x x
(a) (b) (c)

Figure 2.3: Probability mass function: seizures

i. Function (a). 
0.17, if x = 0
P (X = x) =
0.21, if x = 2
ii. Function (b). 
0.18, if x = 4
P (X = x) =
0.11, if x = 6
iii. Function (c). 

 0.17, if x=0
0.21, if x=2




0.18, if x=4

P (X = x) =

 0.11, if x=6
0.16, if x=8




0.17, if x = 10


2. Rolling a pair of dice: number of fours rolled.

Let X be the number of 4’s rolled. Assume the dice are fair.
1 20 25 30
(a) P (X = 2) = P {(4, 4)} = (i) 36
(ii) 36
(iii) 36
(iv) 36
.
(b) Since P (X = 1) =
(i) P {(1, 4), (2, 4), (3, 4), (5, 4), (6, 4)}
(ii) P {(4, 1), (4, 2), (4, 3), (4, 5), (4, 6)}
(iii) P {(1, 4), (2, 4), (3, 4), (5, 4), (6, 4), (4, 1), (4, 2), (4, 3), (4, 5), (4, 6)}
1 10 25 30
(c) Then P (X = 1) = (i) 36
(ii) 36
(iii) 36
(iv) 36
.
(d) P (X = 0) = 1 − P (X = 1) − P (X = 2) =
(i) 11
36
20
(ii) 36 (iii) 25
36
30
(iv) 36 .
(e) The probability distribution of X is
x 0 1 2
25 10 1
P (X = x) 36 36 36

(i) True (ii) False

Section 2. Probability Mass Functions (LECTURE NOTES 3) 35

3. Flipping until a head comes up. A (weighted) coin has a probability of p = 0.7
of coming up heads (and so a probability of q = 1 − p = 0.3 of coming up tails).
Let X be the number of flips until a head comes up or until a total of 4 flips
are made.

(a) Head may come up on first flip, f (1) = P {H} =

(i) 0.3 (ii) 0.7 (iii) 0.3(0.7) (iv) 0.32 (0.7).
(b) Head may come up on second flip, f (2) = P {T H} =
(i) 0.3 (ii) 0.7 (iii) 0.3(0.7) (iv) 0.32 (0.7).
(c) f (3) = P {T T H} =
(i) 0.3 (ii) 0.7 (iii) 0.3(0.7) (iv) 0.32 (0.7).
(d) f (4) = P {T T T T, T T T H} = 1 − p(1) − p(2) − p(3) =
(i) 0.027 (ii) 0.063 (iii) 0.210 (iv) 0.700.
(e) The probability distribution of X is
x 1 2 3 4
f (x) 0.700 0.210 0.063 0.027
(i) True (ii) False
(f) Conditional probability.

P (X ≥ 3 ∩ X ≥ 2) P (X ≥ 3) 0.063 + 0.027
P (X ≥ 3|X ≥ 2) = = = =
P (X ≥ 2) P (X ≥ 2) 1 − 0.7

(i) 0.2 (ii) 0.3 (iii) 0.4 (iv) 0.5.

4. Geometric probability mass function: bull’s eye. There is a 15% (p = 0.15)

chance of hitting a bull’s eye on a dart board. Throws are independent of one
another.

(a) The chance the first bull’s eye occurs on the first try is, of course, 15%.
The chance the first bull’s eye occurs on the second try equals the chance
a miss occurs on the first try and a bull’s eye occurs on the second try,
f (2) = (0.85)0.15 = (i) 0.1155 (ii) 0.1275 (iii) 0.1385 (iv) 0.2515.
(b) The chance the first bull’s eye occurs on the third try is equal to the chance
of two misses and then a bull’s eye occurs on the third try,
f (3) = (0.85)(0.85)0.15 = (0.85)2 0.15 ≈
(i) 0.078 (ii) 0.099 (iii) 0.108 (iv) 0.158.
(c) The chance the first bull’s eye occurs on the fourth try is
f (4) = (0.85)3 0.15 ≈
(i) 0.078 (ii) 0.092 (iii) 0.108 (iv) 0.151.
36 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

(d) The chance the first bull’s eye occurs on the eleventh try, x = 11, is
f (11) = q x−1 p = (0.85)11−1 0.15 ≈
(i) 0.01 (ii) 0.02 (iii) 0.03 (iv) 0.04.
(e) Since the chance the first bull’s eye occurs on the xth try is f (x) = q x−1 p,
then the sum of this series
∞ ∞
X X X p
f (x) = f (x) = q x−1 p = =
x∈R x=1 x=1
1−q

p
(i) p (ii) q (iii) q
(iv) 1.

5. Probability mass function and cumulative distribution function: flipping a coin

twice. Let the number of heads flipped in two flips of a coin be a discrete
random variable X with pmf f (x) is given in table and Figure 2.4(a), and cdf
F (x) = P (X ≤ b) given in Figure 2.4(b).

x 0 1 2
f (x) = P (X = x) 0.25 0.5 0.25

1 1
0.75 0.75
0.50 0.50
0.25 0.25

0 1 2 0 1 2
(a) density (b) distribution

Figure 2.4: Density and distribution function: flipping a coin twice

(a) f (0) = P (X = 0) =
(i) 0 (ii) 0.25 (iii) 0.50 (iv) 0.75.
(b) f (1) = P (X = 1) =
(i) 0 (ii) 0.25 (iii) 0.50 (iv) 0.75.
(c) f (2) = P (X = 2) =
(i) 0 (ii) 0.25 (iii) 0.50 (iv) 0.75.
(d) F (0) = P (X ≤ 0) = P (X = 0) =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.
Section 2. Probability Mass Functions (LECTURE NOTES 3) 37

(e) F (1) = P (X ≤ 1) = P (X = 0) + P (X = 1) =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.
(f) F (2) = P (X ≤ 2) = P (X = 0) + P (X = 1) + P (X = 2) =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.
(g) Also, if b < 0, F (b) = P (X ≤ b) =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.
(h) And, if b ≥ 2, F (b) = P (X ≤ b) =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.
(i) So, in this case, 

 0, b<0
0.25, 0≤b<1

F (b) =
 0.75,
 1≤b<2
1, b ≥ 2.


(i) True (ii) False

(j) The discontinuous “step function” graph of this F (b) is given in Fig-
ure 2.4(b). Notice that F (b) is right continuous, which is indicated by
the solid and empty endpoints on the graph of this distribution function.
The height of a “step” in distribution F (b) is equal to the height of the
corresponding “stick” in density p(b).
(i) True (ii) False
(k) P (X < 1) = P (X = 0) = 0.25 =
(i) F (0) (ii) F (1) (iii) F (2) (iv) F (3).
(l) P (X < 2) = P (X = 0) + P (X = 1) = 0.75 =
(i) F (0) (ii) F (1) (iii) F (2) (iv) F (3).
(m) P (X > 1) = 1 − P (X ≤ 1) = 1 − F (1) = 1 − 0.75 =
(i) 0 (ii) 0.25 (iii) 0.75 (iv) 1.

6. Another discrete cumulative distribution function.

Let random variable X have distribution,


 0, b < 0
 1
3
, 0≤b<1
F (b) = 1
, 1≤b<2
 2


1, b ≥ 2.

(a) f (0) = P (X = 0) =
(i) 0 (ii) 16 (iii) 13 (iv) 12 .
(b) f (1) = P (X = 1) = P (X ≤ 1) − P (X ≤ 0) = F (1) − F (0) =
(i) 0 (ii) 16 (iii) 13 (iv) 12 .
38 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

1 1

1/2 1/2
1/3 1/3
1/6
0 1 2 0 1 2
(a) density (b) distribution

Figure 2.5: pmf and cdf

(c) f (2) = P (X = 2) = P (X ≤ 2) − P (X ≤ 1) = F (2) − F (1) =

(i) 0 (ii) 61 (iii) 13 (iv) 12 .
(d) Random variable X is discrete, not continuous, because the associated
F (b) is a discontinuous (“step”, in this case) function.
(i) True (ii) False
(e) Notice that
(1) limb→−∞ F (b) = 0,
(2) limb→∞ F (b) = 1,
(3) if b1 < b2 , then F (b1 ) ≤ F (b2 ); that is, F is nondecreasing.
(i) True (ii) False
(f) F (1) = P (X ≤ 1) = P (X = 0) + P (X = 1) =
(i) 0 (ii) 61 (iii) 13 (iv) 12 .
(g) P (X < 2) = P (X = 0) + P (X = 1) =
(i) 0 (ii) 16 (iii) 13 (iv) 12 .
(h) P (X ≤ 1.5) = P (X = 0) + P (X = 1) =
(i) 0 (ii) 16 (iii) 13 (iv) 12 .

7. Mode and Median, Discrete.

(a) Number of Ears of Corn. The number of ears of corn, Y , on a typical corn
plant has the following probability distribution.
x 0 2 4 6 8 10
f (x) 0.17 0.21 0.18 0.11 0.16 0.17
The mode number of ears of corn with largest probability,
(i) 0 (ii) 2 (iii) 4 (iv) 6.

The median number of ears of corn, m, is equal to the smallest

number of ears of corn where there is at least a 50% chance of getting less
Section 2. Probability Mass Functions (LECTURE NOTES 3) 39

0.50 less than equal to 1 0.875 greater or equal to1

0.56 less than or equal to 4 0.62 greater than or equal to 4
0.25
0.21 f(x)
0.20 0.18
0.17 3/8 3/8
0.16 0.17
f(x)
0.15 3/8
0.11
0.10 2/8
1/8 1/8
0.05 1/8

0 0
0 2 4 6 8 10 0 1 2 3
x x

mode = 2 median = 4 modes = 1, 2

median = 1

(a) mode and median number of ears of corn (b) mode and median number of swallows

Figure 2.6: Mode and Median for Discrete Probability Distributions

than or equal to this number and also at least a 50% of getting more than
this number), in other words the median m =
i. 2 ears of corn since there is a 17% + 21% = 38% chance of getting
less than or equal to 2 ears and a 21% + 18% + 11% + 16% + 17%
= 83% chance of getting 2 or more.
ii. between 2 and 4 ears of corn since there is a 17% + 21% = 38% chance
of getting less than or equal to 2 ears and a 18% + 11% + 16% + 17%
= 62% chance of getting 4 or more.
iii. 4 ears of corn since there is a 17% + 21% + 18% = 56% chance of
getting less than or equal to 4 ears and a 18% + 11% + 16% + 17%
= 62% chance of getting 4 or more
(b) Swallows. The number of swallows, X, in any group of three birds is given
by the following probability distribution.
X 0 1 2 3
1 3 3 1
P (X = x) 8 8 8 8

The mode number of swallows is the number with largest probability,

(i) 0 (ii) 1 (iii) 2 (iv) 3.

The median number of swallows, m, is smallest number

i. m = 0 because
1 8
P (X ≤ 0) = < 0.5 and P (X ≥ 0) = ≥ 0.5.
8 8
ii. m = 1 because
4 7
P (X ≤ 1) = ≥ 0.5 and P (X ≥ 1) = ≥ 0.5.
8 8
40 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

iii. m = 2 because
7 4
P (X ≤ 2) = ≥ 0.5 and P (X ≥ 2) = ≥ 0.5.
8 8

2.3 The Hypergeometric and Binomial Distribu-

tions
The hypergeometric probability mass function, the probability of selecting n items at
random without replacement from a population of size N with N1 type I objects and
N2 type II objects, N1 + N2 = N , where x of the n objects are type I objects, is
  
N1 N2
x n−x
f (x) = P (X = x) =   , x = 1, . . . , n.
N
n

Random variable X is the number of type I objects from the sample of size n; N1 ,
N2 , N , and n are all parameters of this pmf.

The Bernoulli experiment has the following properties:

(1) An outcome from any Bernoulli experiment trial is either a success or failure,
(2) under repetition, trial outcomes are independent, sampling is with replacement,
(3) the probability of success in any trial is constant, is p.

Assuming a sequence of n Bernoulli trials, each with probability of success p, the

binomial probability mass function is,
 
n
f (x) = P (X = x) = px (1 − p)n−x , x = 0, 1, . . . , n,
x

where random variable X is the number of successes in n trials; n and p are parameters
of this pmf and “X is b(n, p)” means random variable X is a binomial with parameters
n and p.

The hypergeometric can be approximated by the binomial as long as only a small

sample, at most 5%, of the population, n ≤ 0.05N , is selected without replacement,
in which case, let
N1
p= .
N

The unusual-event principle says if we make an assumption and this assumption leads
to only a small chance of an event occurring, the event is unusual, then the assumption
Section 3. The Hypergeometric and Binomial Distributions (LECTURE NOTES 3)41

is probably incorrect. For example, if we assume X is b(n, p), but find out the observed
event of x successes in n trials is unusual because

P (X ≤ x) ≤ 0.05 or P (X ≥ x) ≤ 0.05,

then our assumption X is b(n, p) is possibly incorrect.

Exercise 2.3 (The Hypergeometric and Binomial Distributions)

1. Hypergeometric: sampling red marbles from an urn. Two marbles are taken,
n = 2, one at a time, without replacement, from an urn which has N1 = 6 red
and N2 = 10 blue marbles. We win $2 for each red marble chosen, lose $1 for
each blue marble chosen. Let X be number of red marbles and Y be winnings.

(a) The chance both marbles are red, X = 2, is

     
N1 N2 6 10
x n−x 2 2−2
P {RR} = P (X = 2) =   =   =
N 16
n 2
        

6 
10  
9 
8  
8 
11 
2 0 1 2 1 2
(i)   (ii)   (iii)   = 0.125.

16  
16  
16 
2 3 3
dhyper(2,6,10,2) # hypergeometric pmf

[1] 0.125

(b) Alternatively, P {RR} =

6 5 10 9 12 11 16 15
(i) 16 × 15 (ii) 16 × 15 (iii) 16
× 15
(iv) 16
× 15
= 0.125.
(c) Since the winnings are Y = $4 when both marbles are red,
P (Y = $4) = P {RR} = P (X = 2) =
(i) 0.025 (ii) 0.125 (iii) 0.225 (iv) 0.500.
(d) The chance both marbles are blue, X = 0, is
     
N1 N2 6 10
x n−x 0 2−0
P {BB} = P (X = 0) =   =   =
N 16
n 2
        

6 
10  
9 
8  
6 
10 
2 0 1 2 0 2
(i)   (ii)   (iii)   = 0.375.

16  
16  
16 
2 3 2
42 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

dhyper(0,6,10,2) # hypergeometric pmf

[1] 0.375

(e) Also P {BB} = P (Y = −$2) =

6 5 10 9 12 11 16 15
(i) 16 × 15 (ii) 16 × 15 (iii) 16
× 15
(iv) 16
× 15
= 0.375.
(f) P {RB, BR} = P (Y = $1) = 1 − P {RR} − P {BB} =
(i) 0.025 (ii) 0.125 (iii) 0.225 (iv) 0.500.
(g) The hypergeometric pmf of number of red marbles, X, is
x 0 1 2
P (X = x) 0.375 0.500 0.125
(i) True (ii) False
(h) The hypergeometric pmf of payoffs, Y , is
y -$2 $1 $4
P (Y = y) 0.375 0.500 0.125
(i) True (ii) False

2. Hypergeometric: televisions. Seven television (n = 7) tubes are chosen at ran-

dom from a shipment of N = 240 television tubes of which N1 = 15 are defective.

(a) The probability that X = 4 of the chosen televisions are defective is

  
N1 N2
x n−x
f (4) =   =
N
n
           

15  ×
225  
15  ×
225  
15 ×
225 
4 3 3 3 4 3
(i)   (ii)   (iii)  

240  
240  
240 
4 7 7
(b) The probability X = 4 of the chosen televisions are defective is
f (4) =
(i) 0.0003069 (ii) 0.0005069 (iii) 0.0006069 (iv) 0.0007069.
dhyper(4,15,225,7) # hypergeometric pmf

[1] 0.0003069143

(c) The probability X = 5 of the chosen televisions are defective is

f (5) =
(i) 0.000007069 (ii) 0.000009084 (iii) 0.00010069 (iv) 0.00013059.
dhyper(5,15,225,7) # hypergeometric pmf
Section 3. The Hypergeometric and Binomial Distributions (LECTURE NOTES 3)43

[1] 9.083563e-06

(d) The probability at most X = 5 of the chosen televisions are defective is

P (X ≤ 5) =
(i) 0.900 (ii) 0.925 (iii) 0.950 (iv) 0.999.
phyper(5,15,225,7,lower.tail=TRUE) # hypergeometric cdf

[1] 0.9999999

(e) The probability at least X = 1 of the chosen televisions are defective is

P (X ≥ 1) = 1 − P (X < 1) = 1 − P (X = 0) =
(i) 0.367 (ii) 0.435 (iii) 0.545 (iv) 0.633.
phyper(0,15,225,7,lower.tail=FALSE) # 1 - hypergeometric cdf

[1] 0.3672717

3. Hypergeometric: capture–recapture. To determine approximate number of

perch, N , in Lake Fishalot, n = 45 are captured at random from the lake,
tagged and let go back into the lake. A short while later, another n = 32 perch
are captured, of which x = 2 are found to be tagged. Approximately how many
perch are in Lake Fishalot?

(a) The chance two of the second group of captured fish are tagged is
  
N1 N2
x n−x
f (2) =   =
N
n
           

32  ×
N  
45  ×
N − 45  
45  ×
N 
2 43 2 32 − 2 43 2
(i)   (ii)   (iii)   ,

N  
N  
N 
3200 32 3200
(b) Guess N = 500. In this case, chance two of 32 fish chosen are tagged is
      
N1 N2 45 500 − 45
×
x n−x 2 32 − 2
f (2) =   =   =
N 500
n 32

(i) 0.24 (ii) 0.26 (iii) 0.29 (iv) 0.32.

dhyper(2,45,455,32) # hypergeometric pmf

[1] 0.2405091
44 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

(c) Guess N = 750. In this case, chance two of 32 fish chosen are tagged is
      
N1 N2 45 750 − 45
×
x n−x 2 32 − 2
f (2) =   =   =
N 750
n 32

(i) 0.24 (ii) 0.26 (iii) 0.29 (iv) 0.32.

dhyper(2,45,705,32) # hypergeometric pmf

[1] 0.2853607

(d) Guess N = 1000. In this case, chance two of 32 fish chosen are tagged is
f (2) = (i) 0.24 (ii) 0.26 (iii) 0.29 (iv) 0.32.
dhyper(2,45,955,32) # hypergeometric pmf

[1] 0.2570794

(e) A summary of results are given in the following table.

N 500 750 1000
f (2) 0.24 0.29 0.26
Since the largest chance that 2 of 32 fish chosen are tagged is 0.29, then it
seems from the choices given, the number of fish in Lake Fishalot is
N = (i) 500 (ii) 750 (iii) 1000 (iv) 1250.
(f) The approximation to N would improve if we had more than three f (2)
to choose from; however, more effort would be required in calculating the
extra f (2). Differentiating
   
45 N − 45
×
2 32 − 2
 
N
32

with respect to N and then setting to zero, to locate the maximum N is

also possible, but difficult to do.
(i) True (ii) False

4. Bernoulli: Flipping a coin. The number of heads, X, in one flip of a coin, is

given by the following probability function,

f (x) = (0.25)x (0.75)1−x , x = 0, 1.

(a) The chance of flipping 1 head (X = 1) is

f (1) = (0.25)1 (0.75)1−1 = (i) 0 (ii) 0.25 (iii) 0.50 (iv) 0.75.
(b) This coin is (i) fair (ii) unfair.
Section 3. The Hypergeometric and Binomial Distributions (LECTURE NOTES 3)45

(c) The chance of flipping no heads (X = 0) is

f (0) = (0.25)0 (0.75)1−0 = (i) 0 (ii) 0.25 (iii) 0.50 (iv) 0.75.
(d) A “tabular” version of this probability distribution of flipping a coin is
i. Distribution A.
x 0 1
f (x) 0.25 0.75
ii. Distribution B.
x 0 1
f (x) 0.75 0.25
iii. Distribution C.
x 0 1
f (x) 0.50 0.50
(e) The number of different ways of describing a distribution include (choose
one or more) (i) function (ii) table (iii) graph.

5. Airplane engines. Each engine of four (n = 4) on an airplane fails 11% (p =

0.11, q = 1 − p = 0.89) of the time. Assume this problem obeys the conditions
of a binomial experiment.
(a) The chance
 two
 engines fail is
4
f (2) = 0.112 0.892 = (i) 0.005 (ii) 0.011 (iii) 0.058 (iv) 0.157.
2
dbinom(2,4,0.11) # binomial pmf

[1] 0.05750646

(b) The chance

 three
 engines fail is
4
f (2) = 0.113 0.891 = (i) 0.005 (ii) 0.011 (iii) 0.040 (iv) 0.057.
3
dbinom(3,4,0.11) # binomial pmf

[1] 0.00473836

(c) The chance at mosttwoengines fail is

4
P (X ≤ 2) = 2x=0 0.11x 0.894−x ≈
P
x
(i) 0.991 (ii) 0.995 (iii) 0.997 (iv) 0.999.
pbinom(2,4,0.11) # binomial cdf

[1] 0.9951152

(d) The chance at mostthree

 engines fail is
P3 4
P (X ≤ 3) = x=0 0.11x 0.894−x ≈
x
(i) 0.991 (ii) 0.995 (iii) 0.997 (iv) 0.999.
pbinom(3,4,0.11,lower.tail=TRUE) # binomial cdf
46 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

[1] 0.9998536

(e) The chance at least three engines fail is

P (X ≥ 3) = 1 − P (X ≤ 2) ≈
(i) 0.005 (ii) 0.010 (iii) 0.016 (iv) 0.023.
pbinom(2,4,0.11,lower.tail=FALSE) # 1 - binomial cdf
1 - pbinom(2,4,0.11) # 1 - binomial cdf

[1] 0.00488477

(f) Since
P (X ≤ 3) ≈ 0.999 > 0.05 and P (X ≥ 3) ≈ 0.005 < 0.05,
it (i) is (ii) is not unusual 3 engines fail, assuming 11% engine failure.
6. Binomial distribution: shaded region.
Let 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, where part of the square, x < y, is shaded.

x<y
y

0 1
x

Figure 2.7: Area of x < y

(a) If points are taken at random from this square, determine the probability
of choosing a point from the shaded region, which is half the square.
p = (i) 0.025 (ii) 0.050 (iii) 0.069 (iv) 0.500.
(b) If two points are taken at random from a square, what is the chance both
are from the shaded region? Clearly,
p2 = 0.52 = (i) 0.25 (ii) 0.50 (iii) 0.069 (iv) 0.119
or, using
 binomial
 pmf, since p = 0.5, n = 2, and x = 2, then
2
f (2) = 0.52 0.50 = (i) 0.25 (ii) 0.050 (iii) 0.069 (iv) 0.119.
2
dbinom(2,2,0.5) # binomial pmf

[1] 0.25

(c) If ten points are taken at random from a square, what is the chance seven
are from the shaded region? Using binomial pmf, since p = 0.5, n = 10,
and x =7, then

10
f (7) = 0.57 0.53 = (i) 0.117 (ii) 0.252 (iii) 0.369 (iv) 0.419.
7
Section 3. The Hypergeometric and Binomial Distributions (LECTURE NOTES 3)47

dbinom(7,10,0.5) # binomial pmf

[1] 0.1171875

(d) If n = 11,
 p =0.5, x = 3, then
11
f (3) = 0.53 0.58 = (i) 0.081 (ii) 0.231 (iii) 0.258 (iv) 0.319.
3
dbinom(3,11,0.5) # binomial pmf

[1] 0.08056641

7. Multiple choice questions. On a multiple–choice exam with 4 possible answers

for each of the 5 questions, what is the probability that a student should get at
most 3 correct answers just by guessing?

(a) Since there are five questions,

n = (i) 2 (ii) 3 (iii) 4 (iv) 5.
(b) Since a student wants 3 or more correct answers,
x = (i) 2 (ii) 2, 3 (iii) 2, 3, 4 (iv) 3, 4, 5.
(c) Since the student is choosing from 4 answers at random,
p = (i) 21 (ii) 13 (iii) 41 (iv) 15 .
(d) The chance a student
 should
 get at most 3 correct answers is
P3 5 1 x 3 5−x



P (X ≤ 3) = x=0 4 4
≈
x
(i) 0.697 (ii) 0.704 (iii) 0.812 (iv) 0.984.
pbinom(3,5,0.25) # binomial cdf

[1] 0.984375

(e) The chance a student should get at least 3 correct answers is

P (X ≥ 3) = 1 − P (X ≤ 2) ≈
(i) 0.097 (ii) 0.104 (iii) 0.112 (iv) 0.284.
1 - pbinom(2,5,0.25) # 1 - binomial cdf

[1] 0.1035156

(f) Since

P (X ≤ 3) = 0.984 > 0.05 and P (X ≥ 3) = 0.104 > 0.05,

it (i) is (ii) is not unusual to score 3 correct answers, guessing at random.

8. Lawyer. A lawyer estimates she wins 40% (p = 0.4) of her cases. Assume
each trial is independent of one another and, in general, this problem obeys
the conditions of a binomial experiment. The lawyer presently represents 10
(n = 10) defendants.
(a) The tabular form of this distribution is given by,
48 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

r 0 1 2 3 4 5 6 7 8 9 10
P (R = r) 0.006 0.040 0.121 0.215 0.251 0.201 0.111 0.043 0.011 0.002 0.000

where, for example, the chance of her winning 6 of 10 cases is 0.111. In a

similar way, the chance of her winning 4 of 10 cases is
(i) 0.121 (ii) 0.215 (iii) 0.251 (iv) 0.351.
(b) The graphical form of the probability distribution is given in Figure 2.8.

p(y)

0.40
0.30
0.20

0.10

0 1 2 3 4 5 6 7 8 9 10
number of cases won, y

Figure 2.8: Binomial distribution: lawyer wins

The number of cases the lawyer has most chance of winning is

(i) one (ii) two (iii) three (iv) four.
(c) The binomial distribution, in this case, is
(i) skewed left (ii) skewed right (iii) more or less symmetric .
(d) In general, the binomial distribution will sometimes, but not always, be
symmetric. If her chance of winning was p = 0.1 instead of p = 0.4, the
binomial distribution would be
(i) skewed left (ii) skewed right (iii) more or less symmetric.
(e) Understanding the binomial formula.
Chance of nine wins and then a lose would be p9 q = 0.49 0.61 ,
eight wins, one lose, one win would be: p8 qp = p9 q = 0.49 0.61 ,
seven wins, one lose, two wins would be: p7 qp2 = p9 q = 0.49 0.61 , . . . .
Since nine wins occurs ten different
  ways, chance nine wins would be:
10
10 × p9 q = 10 × 0.49 0.61 = 0.49 0.61 .
9
In
 general,
 chance of x wins in n trials is binomial formula,
n
px q n−x , x = 0, 1, . . . , n
x
(i) True (ii) False
(f) Conditional versus unconditional binomial.
Section 3. The Hypergeometric and Binomial Distributions (LECTURE NOTES 3)49

Since P (X > 3) = 1 − P (X ≤ 3) ≈
(i) 0.46 (ii) 0.58 (iii) 0.62 (iv) 0.67,
1 - pbinom(3,10,0.4) # 1 - binomial cdf

[1] 0.6177194

and
P (X > 4 + 3, X > 4) P (X > 4 + 3) P (X > 7)
P (X > 4+3|X > 4) = = = ≈
P (X > 4) P (X > 4) P (X > 4)

(i) 0.034 (ii) 0.048 (iii) 0.075 (iv) 0.089;

(1 - pbinom(7,10,0.4))/(1 - pbinom(4,10,0.4)) # conditional probability

[1] 0.03350957

in other words, P (X > 3) = 0.62 6= P (X > 4 + 3|X > 4) = 0.03. The

chance the lawyer wins at least 4 cases is not equal to the chance she wins
at least 8 cases, given winning at least 5 cases.

9. Binomial approximation to the hypergeometric: televisions. Seven television

(n = 7) tubes are chosen at random from a shipment of N = 240 television
tubes of which N1 = 15 are defective.

(a) We sample without replacement; that is, every time a TV is chosen, we do

not replace it to be potentially chosen again. In other words, the chance
of choosing a defective TV, every time a TV is chosen, changes or depends
on the number of defective TVs that were chosen before it.
(i) True (ii) False
(b) If the sample size, n, is small relative to the number of televisions, N ,
n < 0.05N , say, the hypergeometric can be approximated by a binomial.
The chance, p = NN1 , of choosing a defective TV, every time a TV is chosen,
does not change “that much” when Nn < 0.05. Since Nn = 240 7
≈ 0.029 <
0.05, the binomial will probably approximate the hypergeometric
(i) very closely. (ii) somewhat closely. (iii) not closely at all.
(c) (Exact) hypergeometric. Probability x = 4 of n = 7 chosen TVs are
defective is
      
N1 N2 15 225
×
x n−x 4 3
f (4) =   =   =
N 240
n 7

(i) 0.0003069 (ii) 0.0004400 (iii) 0.0006069 (iv) 0.0007069.

dhyper(4,15,225,7) # exact hypergeometric pmf

[1] 0.0003069143
50 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

(d) Approximate binomial. Since p = NN1 = 24015

= 0.0625 and so a binomial
approximation to probability x = 4 of n = 7 chosen TVs are defective is:
 
7
f (4) = 0.06254 (1 − 0.0625)7−4 =
4

(i) 0.0003069 (ii) 0.0004400 (iii) 0.0006069 (iv) 0.0007069.

dbinom(4,7,0.0625) # approximate binomial pmf

[1] 0.0004400499

10. Multinomial: faculty and subjects. There are 9 different faculty members and
3 subjects: mathematics, statistics and physics. There is a 50%, 35% and 15%
chance a faculty member teaches mathematics, statistics and physics, respec-
tively.

(a) Chance 4, 3 and 2 faculty members teach mathematics, statistics and

physics, respectively, is
9!
p(4, 3, 2) = 0.54 0.353 0.152 ≈
4!3!2!
(i) 0.055 (ii) 0.067 (iii) 0.076 (iv) 0.111
dmultinom(c(4,3,2), prob = c(0.5,0.35,0.15)) # multinomial pmf

[1] 0.07596914

(b) Chance 4, 4 and 1 faculty members teach mathematics, statistics and

physics, respectively, is
9!
p(4, 4, 1) = 0.54 0.354 0.151 ≈
4!4!1!
(i) 0.089 (ii) 0.098 (iii) 0.108 (iv) 0.131
dmultinom(c(4,4,1), prob = c(0.5,0.35,0.15)) # multinomial pmf

[1] 0.08863066

2.4 The Poisson Distribution

Poisson probability mass function, used to describe probabilities of count, X, of
occurrences (successes) of an event in a specified time (or space) interval, is
λx −λ
f (x) = P (X = x) = e , x = 0, 1, . . . ,
x!
where λ > 0 is average number of occurrences of an event; also, random variable X
obeys conditions of Poisson process:
Section 4. The Poisson Distribution (LECTURE NOTES 3) 51

• zero probability of two or more successes in sufficiently small subinterval,

• probability of success same if two intervals are of equal length,

• number successes in one interval independent of number of successes in any

other nonoverlapping interval.

If n ≥ 100, np ≤ 100, Poisson pmf can be used to approximate binomial pmf where
λ = np.

Exercise 2.4 (The Poisson Distribution)

1. Poisson: number of photon hits.

Piece of iron bombarded with electrons releases a number of photons, X, to a
surrounding magnetic detection field. An average of λ = 5 particles hit magnetic
detection field per (one) microsecond. Assume this is a Poisson process.

(a) Chance x = 0 photons hit field in one microsecond

λx −λ 50 −5
f (0) = e = e ≈
x! 0!
(i) 0.007 (ii) 0.008 (iii) 0.009.
dpois(0,5) # Poisson pmf

[1] 0.006737947

(b) Chance x = 2 photons hit field in one microsecond

λx −λ 52 −5
f (2) = e = e ≈
x! 2!
(i) 0.06 (ii) 0.07 (iii) 0.08.
dpois(2,5) # Poisson pmf

[1] 0.08422434

(c) Chance at most x = 2 photons hit field in one microsecond

50 −5 51 −5 52 −5
P (X ≤ 2) = e + e + e ≈
0! 1! 2!
(i) 0.08 (ii) 0.12 (iii) 0.18.
ppois(2,5) # Poisson cdf

[1] 0.124652
52 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

(d) Chance at least x = 2 photons hit field in one microsecond

52 −5 53 −5
P (X ≥ 2) = e + e + · · · forever
2! 3!
= 1 − P (X < 2) = 1 − P (X ≤ 1)
= 1 − [P (0) + P (1)]
 0
5 −5 51 −5

= 1− e + e ≈
0! 1!

(i) 0.91 (ii) 0.93 (iii) 0.96.

1 - ppois(1,5) # 1 - Poisson cdf
ppois(1,5,lower.tail=FALSE) # 1 - Poisson cdf

[1] 0.9595723

(e) Chance x = 2 photons hit field in two microseconds

(5 · 2)2 −5·2
P (2) = e ≈
2!
(i) 0.002 (ii) 0.005 (iii) 0.006.
dpois(2,10) # Poisson pmf, 2 x 5 = 10

[1] 0.002269996

(f) Chance at most x = 3 photons hit field in two microseconds

P (X ≤ 3) ≈

(i) 0.01 (ii) 0.06 (iii) 0.08.

ppois(3,10) # Poisson cdf, 2 x 5 = 10

[1] 0.01033605

(g) Chance at least x = 21 photons hit field in four microseconds

P (X ≥ 21) = 1 − P (X < 21) = 1 − P (X ≤ 20) ≈

(i) 0.44 (ii) 0.52 (iii) 0.68.

1 - ppois(20,20) # 1 Poisson cdf, 4 x 5 = 20

[1] 0.4409074

(h) Poisson process? Match columns.

Poisson process photon example

(a) zero chance more than one success in small subinterval (A) assume “zero” chance two hits at same time
(b) chance of success same if two intervals are of equal length (B) assume chance number hits same per microsecond
(c) independence of successes in different intervals (C) assume hits independent of one another

Poisson process (a) (b) (c)

photon example
Section 4. The Poisson Distribution (LECTURE NOTES 3) 53

2. Poisson: Bryozoan count.

During a biology study, 250 bryozoans are found attached on a submerged 10
centimeter by 10 centimeter (100 centimeters2 ) plate, an average of λ = 250
100
=
2.5 bryozoans per one centimeter2 . Assume this is a Poisson process.

(a) Chance x = 0 bryozoans attached in one centimeter2

λx −λ 2.50 −2.5
f (0) = e = e ≈
x! 0!
(i) 0.07 (ii) 0.08 (iii) 0.09.
dpois(0,2.5) # Poisson pmf

[1] 0.082085

(b) Chance at most x = 2 bryozoans attached in one centimeter2

P (X ≤ 2) ≈

(i) 0.54 (ii) 0.62 (iii) 0.78.

ppois(2,2.5) # Poisson cdf

[1] 0.5438131

(c) Chance between x = 2 and x = 4 bryozoans in one centimeter2 , inclusive:

P (2 ≤ X ≤ 4) = P (X ≤ 4) − P (X < 2) = P (X ≤ 4) − P (X ≤ 1) ≈

(i) 0.60 (ii) 0.73 (iii) 0.76.

ppois(4,2.5) - ppois(1,2.5) # between Poisson cdfs

[1] 0.6038805

(d) Chance between x = 2 and x = 4 bryozoans in three centimeters2 , inclusive:

P (2 ≤ X ≤ 4) = P (X ≤ 4) − P (X < 2) = P (X ≤ 4) − P (X ≤ 1) ≈

(i) 0.13 (ii) 0.23 (iii) 0.36.

ppois(4,7.5) - ppois(1,7.5) # between Poisson cdfs, 3 x 2.5 = 7.5

[1] 0.1273606

3. Poisson approximation of the binomial: photons. What is the chance x = 15

particles hit in one microsecond? The Poisson distribution can be used to
approximate the binomial distribution by letting λ = np. This is a fairly good
approximation if n ≥ 100 and np ≤ 100.
54 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

(a) Assume n = 2000 particles are released by iron per microsecond, and
there is a chance p = 0.005 that a particle hits the surrounding field per
microsecond. Since n = 2000 and p = 0.005, then n = 2000 > 100 and
np = 2000(0.005) = 10 ≤ 100 and so the conditions for approximation are
satisfied (ii) violated.

(Exact) binomial
2000!
f (15) = × (0.005)15 × (0.995)1985 ≈
15!(2000 − 15)!

(i) 0.00234 (ii) 0.0346 (iii) 0.0445 (iv) 0.0645.

dbinom(15,2000,0.005) # exact binomial pmf

[1] 0.03463059

Approximate Poisson.
since λ = np = 2000(0.005) = (i) 5 (ii) 10 (iii) 15 (iv) 20, then

1015 −10
f (15) = e ≈
15!
(i) 0.00234 (ii) 0.0347 (iii) 0.0445 (iv) 0.0645
dpois(15,10) # approximate Poisson pmf

[1] 0.03471807

(b) If n = 1500 and p = 0.01, then n = 1500 > 100 and

np = 1500(0.01) = 15 ≤ 100 and so the conditions for approxima-
tion are satisfied (ii) violated.

(Exact) binomial
1500!
f (15) = × (0.01)15 × (0.99)985 ≈
15!(1500 − 15)!

(i) 0.1024 (ii) 0.1030 (iii) 0.1245 (iv) 0.1345.

dbinom(15,1500,0.01) # exact binomial pmf

[1] 0.1029519

Approximate Poisson
since λ = np = 1500(0.01) = (i) 5 (ii) 10 (iii) 15 (iv) 20, then

1515 −15
f (15) = e ≈
15!
(i) 0.1024 (ii) 0.1030 (iii) 0.1345 (iv) 0.1445
dpois(15,15) # approximate Poisson pmf
Section 4. The Poisson Distribution (LECTURE NOTES 3) 55

[1] 0.1024359

4. Poisson process: photons. A random variable has a Poisson distribution if the

Poisson process conditions are satisfied. For the photons example, this would
mean the following assumptions are satisfied.
• P (no “hit” occurs in time subinterval)= 1 − p
• P (one “hit” occurs in time subinterval)= p
• P (two or more “hits” occurs in time subinterval)= 0
• Occurrence of hit in each time subinterval is independent of occurrence of
event in other nonoverlapping subintervals.

hit hit miss miss miss miss hit

1 2 3 4 5 6 7
time subintervals

Figure 2.9: Poisson process: sequence of photon hits

(a) The Poisson process assumes all time subintervals are created so small one
and only one photon could hit the magnetic detection field during each
subinterval. During seven time subintervals shown in Figure 2.9, three
hits occur in time subintervals 1, 2 and 7 and four misses occur in time
subintervals 3, 4, 5 and 6.
(i) True (ii) False
(b) The Poisson process assumes the chance a photon hits the magnetic de-
tection field during any one time subinterval is p and the chance it misses
is q = 1 − p. It is impossible to have more than one photon hit the mag-
netic detection field during any time subinterval. Consequently, the time
subintervals are infinitesimally small.
(i) True (ii) False
(c) The Poisson process assumes a hit in each time subinterval is independent
of any other hit in other nonoverlapping subintervals.
(i) True (ii) False
(d) If there was a 20% chance, p = 0.2, a photon hit the magnetic detection
field during any time subinterval then the chance of y = 3 hits in the n = 7
time subintervals is given by the binomial,
   
n x n−x 7
f (3) = p q = 0.23 0.87−3 ≈
x 3
(i) 0.089 (ii) 0.115 (iii) 0.124 (iv) 0.134.
56 Chapter 2. Discrete Random Variables (LECTURE NOTES 3)

dbinom(3,7,0.2) # binomial pmf

[1] 0.114688

(e) If a large number of time subintervals are considered; in other words, as, n
gets bigger, in fact, as n → ∞, it becomes increasingly difficult to calculate
the probability of a number of hits using the binomial. In this case, the
poisson is used instead.
(i) True (ii) False
(f) In general,
λx −λ
 
n
lim px q n−x = e
n→∞ x x!
where λ = np.
(i) True (ii) False

5. Poisson: stars on tiles.

Twenty-four (24) stars are placed on a tiled floor with 90 tiles and it is found,
for example, 70 of tiles had no (0) stars, 17 tiles had 1 star and so on. Does
this distribution of stars approximately follow a Poisson distribution?

x 0 1 2 3
frequency 70 17 2 1

Figure 2.10: Stars on tiles

(a) Fill in the blank for the relative frequency of stars per tile table:
Section 4. The Poisson Distribution (LECTURE NOTES 3) 57

x 0 1 2 3
frequency 70 17 2 1
70
relative frequency 90
≈ 0.78 0.19 0.02
(b) Average number of stars per tile is
24
λ=
90
(i) 0.21 (ii) 0.24 (iii) 0.27 (iv) 0.29.
(c) Assuming Poisson where λ = 0.27, chance x = 0 stars on one tile is

λx −λ 0.270 −0.27
f (0) = e = e ≈
x! 0!
(i) 0.01 (ii) 0.02 (iii) 0.19 (iv) 0.76.
dpois(0,0.27) # Poisson pmf

[1] 0.7633795

(d) Fill in the blank for the Poisson pmf f (x) of stars per tile table:
x 0 1 2 3
frequency 70 17 2 1
70
relative frequency 90
≈ 0.78 0.19 0.02 0.01
f (x) 0.76 0.03 0.00
The Poisson pmf (i) is (ii) is not a close approximation to the relative
frequency of stars per tiles.
dpois(0:3,0.27) # Poisson pmf for 0,1,2,3

[1] 0.763379494 0.206112463 0.027825183 0.002504266