BIOSTATISTICS I

EPIB 650
Module 3: Discrete
Distributions
Professor Yan Li
yli6@[Link]

1
Review

o Probability = Study of randomness

o 0  P(A)  1 for any event A
o P(Ω) = 1, P() = 0
o A’s complement Ā, and P(Ā) = 1 − P(A)
o Mutually exclusive
oP(A ∩ B) = 0
o Mutually independent
oP(A ∩ B) = P(A) × P(B)

2
Review

o Addition law of probability

o P(A ∪ B) = P(A) + P(B) − P(A ∩ B)
o Multiplication law of probability (for mutually
independent events, A1, A2, . . . , Ak )
o P(A1∩ A2 ∩. . . ∩ Ak) = P(A1) × P(A2) × . . . × P(Ak)

3
Review

• Conditional Probability:
P( A  B)
P( A | B) =
P( B)
• If A and B are independent,
oP(A|B) = P(A)
oP(B|A) = P(B)

4
Medical decision-making

• The clinical performance of a laboratory test can be

described in terms of diagnostic accuracy or the ability to
correctly classify subjects into clinically relevant
subgroups
True (disease) status
Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total a+c b+d n

5
 Evaluation of a diagnostic test: given
6
disease status (true)
True (disease) status
Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total Total a + ca+c b + d b+d n

o Sensitivity (True positive rate) = P(T + | D +)= a/(a+c)

o Specificity (True negative rate) = P(T − | D −)= d/(b+d)
 Evaluation of a diagnostic test : given
7
disease status (false)
True (disease) status
Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total Total a + ca+c b + d b+d n

o False Negative Rate = 1-sensitivity = P(T − | D+)= c/(a+c)

o False Positive Rate = 1-specificity = P(T+ | D −)= b/(b+d)
 Prevalence P(D+)
8

True (disease) status

Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total Total a + a+c
c b + d b+d n
 Predictive value : given test results
9

True (disease) status

Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total Total a + a+c
c b + d b+d n

o Positive Predictive Value (PPV) = P(D + |T + )= a/(a+b)

o Negative Predictive Value (NPV) = P(D − |T −)=d/(c+d)
 Receiver operating characteristic (ROC) curve
Cut point
True (disease) status
Disease (D+) Disease (D−) Total
Test positive (T+) a b a+b
Test negative (T−) c d c+d
Total a+c b+d n

o True positive rate = sensitivity = a/(a+c)

o False Positive Rate = 1- specificity = b/(b+d)

10
o We have defined probability and learned some basic tools in
working with probabilities
o How real problems can be put in a probability framework?
o How to assess the probabilities of certain event from actual
data?
1. Discrete probability distribution (Chapter 4)
2. Continuous probability distribution (Chapter 5)

11
Chapter 4
Discrete Probability
Distributions
Definition
• A random variable is a variable whose value is a
numerical outcome of an event
• A discrete random variable X is a random variable
which takes from a discrete set of values (finite
values) with specified probability for each value,
={x1 , x2 ,…, xk}

17
Examples – discrete variables

oThe number of boys in 100 families that each family only has one
child
oThe number of women with positive test results from 5000 women
taken the screening test for breast cancer

18
Example

oReal event: Surgery on three people success (s) or failure (f)

oAll possible outcomes Ω = {sss, ssf, sfs, fss, sff, fsf, ffs, fff}
oDefine X = total number of success(s)
X = 0, {fff}
X = 1, {sff, fsf, ffs}
X = 2, {ssf, sfs, fss}
X = 3, {sss}

Question: Expected total number of surgery success(es)

among the three people?

19
Probability distribution function
• The discrete probability distribution function (probability
mass function: pmf) of X is the assignment of probabilities
to the values of the random variable X
Values of X Probability k

x1 p1
0  pi  1, p
i =1
i =1
x2 p2
. .
. .
. .
xk pk

20
Probability distribution function
• The discrete probability distribution function (probability
mass function: pmf) of X is the assignment of probabilities
to the values of the random variable X
• In the example,
Values of X Probability
x1 = 0, {fff} p1 = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8
x3 = 2, {ssf, sfs, p3 = 3/8
fss}
x4 = 3, {sss} p4 = 1/8

21
Probability distribution function
• The discrete probability distribution function (probability
mass function: pmf) of X is the assignment of probabilities
to the values of the random variable X
Values of X Probability PDF = f(X)
x1 = 0, {fff} p1 = 1/8 f(x1) = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8 f(x2) = 3/8
x3 = 2, {ssf, sfs, fss} p3 = 3/8 f(x3) = 3/8
x4 = 3, {sss} p4 = 1/8 f(x4) = 1/8

22
Cumulative distribution function

• The cumulative distribution function (CDF) of X is

F(x) = Pr(X ≤ x), −∞ < x < +∞

23
 Cumulative distribution function

• The cumulative distribution function (CDF) of X is

F(x) = Pr(X ≤ x), −∞ < x < +∞

Values of X Probability PDF = f(X) CDF=F(X)

x1 = 0, {fff} p1 = 1/8 f(x1) = 1/8 F(x1) = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8 f(x2) = 3/8 F(x2) = 1/8+3/8
x3 = 2, {ssf, sfs, fss} p3 = 3/8 f(x3) = 3/8 F(x3) = 1/8+3/8+3/8
x4 = 3, {sss} p4 = 1/8 f(x4) = 1/8 F(x4) = 1/8+3/8+3/8+1/8=1

24
 Cumulative distribution function
Values of X Probabilit PDF = f(X) CDF=F(X)
y
x1 = 0, {fff} p1 = 1/8 f(x1) = 1/8 F(x1) = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8 f(x2) = 3/8 F(x2) = 1/8+3/8
x3 = 2, {ssf, sfs, p3 = 3/8 f(x3) = 3/8 F(x3) = 1/8+3/8+3/8
fss}
x4 = 3, {sss} p4 = 1/8 f(x4) = 1/8 F(x4) = 1/8+3/8+3/8+1/8=1

1/8 +3/8 +3/8+1/8=1

1/8 +3/8 +3/8

1/8 +3/8

1/8
0 1 2 3 4
25
Cumulative distribution function

Note that
1) F(x) in non-decreasing
2) F(−∞) = 0 and F(+∞) = 1
3) Step function

26
Mean of a discrete random variable
(measure of center)
• Expected value of a discrete random variable, denoted by
μX or E(X) = “average” value of the random variable
k
Values of X Probability 0  pi  1, p i =1
x1 p1 i =1

x2 p2 E ( X ) =  X = x1 p1 + x2 p2 + L + xk pk
. .
. .
. .
xk pk

27
Mean of a discrete random variable
• Expected value of a discrete random variable, denoted by
μX or E(X) = “average” value of the random variable

Values of X Probability k

x1 p1
0  pi  1, p
i =1
i =1
x2 p2 E ( X ) =  X = x1 p1 + x2 p2 + L + xk pk
. .
𝑘
. .
𝐸 𝑋 = 𝜇𝑥 = ෍ 𝑥𝑖 𝑃 𝑋 = 𝑥𝑖
. .
𝑖=1
xk pk

28
Mean of a discrete random variable

Values of X Probability PDF = f(X)

x1 = 0, {fff} p1 = 1/8 f(x1) = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8 f(x2) = 3/8
x3 = 2, {ssf, sfs, fss} p3 = 3/8 f(x3) = 3/8
x4 = 3, {sss} p4 = 1/8 f(x4) = 1/8

1 3 3 1 12
E ( X ) = 0  + 1 + 2  + 3  = = 1.5
8 8 8 8 8

29
Variance of a discrete RV (measure of spread)

• Expected value of a discrete random variable, denoted by

μX or E(X) = “average” value of the random variable

Values of X Probability 𝑘
x1 p1 0 ≤ 𝑝𝑖 ≤ 1, ෍ 𝑝𝑖 = 1
x2 p2 𝑖=1
. . 𝑉𝑎𝑟 𝑋 = 𝜎𝑋2 = 𝑥1 − 𝜇𝑥 2 𝑝1 + ⋯ + 𝑥𝑘 − 𝜇𝑥 2 𝑝𝑘
𝑘
. . 𝑉𝑎𝑟(𝑋) = ෍ 𝑥𝑖 − 𝜇𝑥 2 𝑃(𝑋 = 𝑥𝑖 )
. . 𝑖=1
xk pk

30
Variance of a discrete RV (measure of
spread)
Values of X Probability PDF = f(X)
x1 = 0, {fff} p1 = 1/8 f(x1) = 1/8
x2 = 1, {sff, fsf, ffs} p2 = 3/8 f(x2) = 3/8
x3 = 2, {ssf, sfs, fss} p3 = 3/8 f(x3) = 3/8
x4 = 3, {sss} p4 = 1/8 f(x4) = 1/8

1 3 3 1
Var ( X ) = ( 0 − 1.5 )  + (1 − 1.5 )  + ( 2 − 1.5 )  + ( 3 − 1.5 )  = 0.75
2 2 2 2

8 8 8 8

31
The standard deviation
 The standard deviation σX of X is the square root of
the variance

32
The standard deviation
 The standard deviation σX of X is the square root of
the variance

Var ( X ) = 0.75 = 0.866

33
Equality in distribution

• If X and Y have the same CDFs, we say that they have the same
distribution (they are equal in distribution), i.e.
o Pr(X ≤ a) = Pr(Y ≤ a) for all a

34
Some rules for means

• If X is a random variable and a and b are fixed

numbers, then
oμa+bX = E(a + b·X) = a + b·E(X) = a + b· μX

• If X and Y are random variables, then

oμX+Y = E(X + Y) = μX + μY

35
Some rules for variances
 If X is a random variable and a and b are fixed
numbers, then
o 2a+bX = Var(a + b·X) = b2· Var(X) = b2 2X

 If X and Y are 2 independent random variables, then

o 2X+Y = Var(X + Y) = 2X + 2Y

o 2X−Y = Var(X − Y) = 2X + 2Y

36
Exercise
• Suppose X1,X2, ..., Xn are independent random variables
having the same distribution
• Suppose E(Xi) = μ and Var(Xi) = σ2
• Then

37
Exercise
• In general

38
Exercise
• Suppose X1,X2, ..., Xn are independent random variables
having the same distribution with E(Xi) = μ and Var(Xi) = σ2
𝐸 𝑋1 +𝐸 𝑋2
• =𝜇
2
𝑉𝑎𝑟 𝑋1 +𝑉𝑎𝑟 𝑋2 𝜎2
• =
4 2
𝐸 𝑋1 +𝐸 𝑋2 +𝐸 𝑋3
• =𝜇
3
𝑉𝑎𝑟 𝑋1 +𝑉𝑎𝑟 𝑋2 +𝑉𝑎𝑟 𝑋3 𝜎2
• =
9 3

𝐸 𝑋1 +⋯+𝐸 𝑋𝑛
• 𝐸 𝑋ത = =𝜇
𝑛
𝑉𝑎𝑟 𝑋1 +⋯+𝑉𝑎𝑟 𝑋𝑛 𝜎2
• 𝑉𝑎𝑟 𝑋ത = =
𝑛2 𝑛

39
Special Discrete Distributions
1. Bernoulli distribution
2. Binominal distribution
3. Poisson distribution
Bernoulli Distribution
• A random experiment with outcomes that can be
classified into two categories (also called a
“Bernoulli trial”), for example,
oDisease positive or negative
oSuccess or failure
oAbsent or present
• Oftentimes, a random variable X (called a Bernoulli
RV) is defined to be
o=1 if the trial results in success
o=0 if the trial results in failure

41
Mean and Variance

• The probability distribution of a Bernoulli random

variable X is
oPr(X = 1) =p
oPr(X = 0) = 1 − p = q
• Then the mean and variance of X are
oμ = E[X] = σ2𝑖=1 𝑥𝑖 𝑝(𝑋 = 𝑥𝑖 ) = 1 × 𝑝 + 0 × 1 − 𝑝 = p
oσ2 = Var[X] = σ2𝑖=1 𝑥𝑖 − 𝜇 2 𝑝(𝑋 = 𝑥𝑖 ) = 1 − 𝑝 2 × 𝑝 +
0 − 𝑝 2 × 1 − 𝑝 = p×q

42
Example of a Bernoulli Trial
• Let X be a random variable that represents smoking
status
oX = 1 if person is a smoker
oX = 0 if person is a non-smoker
• Suppose we know that 29% of adults in the U.S. are
smokers
oP(X = 1) = 0.29
oP(X = 0) = 0.71

43
Example of a Bernoulli Trial

• Let X be a random variable that represents smoking

status
oX = 1 if person is a smoker
oX = 0 if person is a non-smoker
• Suppose we know that 29% of adults in the U.S. are
smokers
oP(X = 1) = 0.29
oP(X = 0) = 0.71
• Then the mean and variance of X are
oμ = E[X] = p =0.29
oσ2 = Var[X] = p×q =0.29×0.71=0.2059

44
Repeated Bernoulli Trial

oSuppose the experiment is to select three

individuals and record their smoking status
oLet X denote the number of smokers in the three:
X = 0, 1, 2, 3 are possible outcomes

45
Repeated Bernoulli Trial

oSuppose the experiment is to select three

individuals and record their smoking status
oLet X denote the number of smokers:
X = 0, 1, 2, 3 are possible outcomes
oWhat is the probability distribution function of X?

46
o Yi denotes the smoking status of the ith person
(independent for all i = 1, 2, 3)
o P(Yi =1)=p & P(Yi =0)=1–p = q
o X = Y1 + Y2 + Y3

o What is the PDF of X?

47
Binomial distribution (P(X=x) for x=0,1,2,3)

48
Binomial distribution
o In general, suppose we have n independent
Bernoulli trials with the same probability of
success p on each trial
o Let X denote the number of successes in the n
trials
n
X =  Yi
i =1

o Then the probability distribution function of X is

called the Binomial distribution function with
parameters p (q=1-p) and n
o X~bin(n, p)

49
 Probability mass function
• X is called the Binomial distribution function with
parameters p (q=1-p) and n
oX~bin(n, p)
n k n−k  n  k n−k
P( X = k ) =   p (1 − p ) =   p q , k = 0,1,..., n
k  k 
oBinomial coefficient: number of ways to choose k from n distinct objects without regard to order
n n!
 = ,
 k  (n − k )!k !
k ! = k (k − 1)(k − 2)... 1, 0! = 1

50
Example (defective gene)

oOn average, 1% of the population are defective in gene H

oA random sample of 10 is selected
oWhat is the probability that this sample contains exactly 2
individuals with defective gene H?

51
Solution
• The probability of selecting 2 defectives, then 8 non-
defectives, in a particular order, is (due to independence)
oP(D D N N N N N N N N) = (0.01)2(0.99)8
Any other order of 2 defectives + 8 non-defectives, can be...
1. D N D N N N N N N N
2. D N N D N N N N N N
3. D N N N D N N N N N
4. D N N N N D N N N N
...... 10 
possible different ordering is  
2

52
Solution
• The probability of selecting 2 defectives, then 8 non-
defectives, in a particular order, is (due to independence)
oP(D D N N N N N N N N) = (0.01) 2(0.99)8

• Therefore, the probability of getting 2 defective and 8 non-

defectives in any order is 0.0415

10  10 − 2 10!
P(2 defective) =   0.01 (1 − 0.01)
2
= (0.01) 2 (0.99)8
2 2!8!

53
Mean and Variance
 X~bin(n, p)

Mean and variance of X are

o μ(X) = σ𝑛𝑖=0 𝑥𝑖 𝑃(𝑋 = 𝑥𝑖 ) = ⋯ = np
o σ2(X)= σ𝑛𝑖=0 𝑥𝑖 − 𝑛𝑝 2 𝑃 𝑋 = 𝑥𝑖 = ⋯ = npq

 For example,
o If n = 100 and p = 1/4, the mean is np = 100×(1/4) = 25
o We expect 25 successes

54
 Remark
oIn practice the success probability p in Bernoulli
trial is usually unknown (called a ‘parameter’)
oFor example we often do not know the exact
proportion p of the defectives in a certain population
oWe want to estimate p

55
 Remark
 Estimate p by ‘sample mean’

n
(μ = E[X] = p) ˆ
Xi
P= i =1
n
To this end, we perform the Bernoulli trials
independently n times
o For all i = 1, . . . , n
Xi = 0 if the ith trial is a failure
Xi = 1 if the ith trial is a success

56
Remark

• The experiment, X1, X2 , . .. , Xn are mutually

independent random variables, therefore
n
Xi
pˆ = 
i =1 n

is also a random variable

57
oAfter the experiment Xi results in a given number (0 or
1), denoted as xi
n n
Xi xi
 = 
o i =1 n results in a known number — the
i =1 n

relative frequency of success

n
xi
o  n = an estimate of p
i =1

58
Note
oIn general, an estimator is a random variable
before the experiment (sampling)
oAn estimate is an observation of that random
variable
oThe application of binomial distribution requires
the independence of the n experiment and that
the probability of success (p) is the same for each
of the n experiments

59
Using R
oEvery distribution in R has four functions
oThere is a root name, for example, the root name for the
binomial distribution is binom
oThis root is prefixed by one of the letters
1. p for "probability", the cumulative distribution function
(c. d. f.)
2. q for "quantile", the inverse c. d. f.
3. d for "density", the density function (p. f. or p. d. f.)
4. r for "random", a random variable having the specified
distribution

60
61
62
PDF

• The probability of selecting 2 defectives, then 8 non-

defectives, in a particular order, is (due to independence)
oP(D D N N N N N N N N) = (0.01) 2(0.99)8

10
o P(X = 2 defectives) = 0.012 0.998
2
odbinom(2, 10, 0.01) = 0.00415

63
CDF

• Look up P(X <= 27) when X is has the Bin(100, 0.25) distribution
opbinom(27, size=100, prob=0.25)
opbinom(27, 100, 0.25)

64
 Poisson distribution

oDescribes the occurrence of relatively rare events (in time

or space)
o Probability mass function: X ~ Poisson(𝜇)
A discrete random variable X is said to have a Poisson distribution
with parameter 𝜇 if the probability of x events occurring in a time
period t (or over an area) with rate 𝜆 is equal to
( t ) x −  t
Pr( X = x) = e , x = 0,1, 2,...
x!

where e2.71828 and 𝜇 = 𝜆𝑡

65
Mean and variance
• The mean and variance of X ∼ Poisson () are
E(X) = Var(X) = λt = 
oλ: the expected number of events per unit time (intensity)
o : the expected number of events over the time period t

( t ) k −  t  k − 
Pr( X = k ) = e = e , k = 0,1, 2,...
k! k!

66
Examples
We should think about the Poisson distribution for any
situation that involves counting events

o# of serious accidents per year in a factory

o# of deaths of employees in a hospital per year
o# of white blood cells found in a cubic centimeter of
blood

67
Infection example

oThe infection rate at a Neonatal Intensive Care Unit

(NICU) is typically expressed as a number of infections
per patient days
oThis is obviously counting a number of events across
both time and patients
oDoes this data follow a Poisson distribution?

68
 Assumptions
oRare event?
oSeveral assumptions need to be checked before using
Poisson distribution
✓Note: the term “interval” refers to either a time interval or an
area depending on the context of the problem

69
Assumption 1
1. The probability of observing a single event over a small
interval is approximately proportional to the size of that
interval

Infection example:
1. The probability of getting an infection over a short time
period is proportional to the length of the time period:
In other words, a patient who stays one hour in the NICU
has twice the risk of a single infection as a patient who
stays 30 minutes

70
Assumption 2
2. The probability of two events occurring in the same
narrow interval is negligible

Infection example:
2. For a small enough interval, the probability of getting
two infections is negligible

71
Assumption 3
3. The probability of an event within a certain interval does
not change over different intervals

Infection example:
3. The probability of infection does not change over time
or over infants
o Each infant is equally likely to get an infection over the
same time interval
o For a single infant, the probability of infection early in
the NICU stay is the same as the probability of
infection later in the NICU stay

72
 Assumption 4
4. The probability of an event in one interval is independent
of the probability of an event in any other non-overlapping
interval

Infection example:
4. Independence: Here independence means two things
o The probability of seeing an infection in one child does not
increase or decrease the probability of seeing an infection in
another child
o We also need to that if an infant who gets an infection during one
time interval, it doesn't change the probability that he or she will
get another infection during a later time interval

73
Infection example
• Does this data follow a Poisson distribution?
• All of these assumptions are suspect, but especially the
last two
oThe infection rate per patient changes from early to later in the
stay, since older infants have better immune systems
oOne infected infant increases the chance that other infants will
get the same infection
oSome infants are more infection prone than others

74
Note

oWe should examine all of these assumptions carefully,

especially the last two
oIf either of these last two assumptions are violated, they
can lead to extra variation, sometimes referred to as
overdispersion

75
Note
oX ~ Poisson()
𝜇𝑘 −𝜇
Pr 𝑋 = 𝑘 = 𝑒 , 𝑘 = 0,1,2 …
𝑘!
 k=0 1 2 3
0.1 0.905 0.090 0.005 0.000

 k=0 1 2 3 4 5
0.5 0.607 0.303 0.076 0.013 0.002 0.000

 k=0 1 2 3 4 5 6 7 8
1.5 0.223 0.335 0.251 0.126 0.047 0.014 0.004 0.001 0.000

76
 = 2.5

 = 7.5

 = 15

77
Remark

oPoisson data tends to have distribution that is

skewed to the right
oIt becomes closer to symmetric as the mean of
the distribution increases
oIf your data comes from a Poisson distribution,
then the mean and the variance of your data
should be roughly equal

78
Empirical tests

There are also some empirical ways of checking for a

Poisson distribution:
• Check variance and mean
oThe simplest and handiest way is to see if the variance is
roughly equal to the mean for your Poisson data
• Use a histogram
oA histogram of the Poisson data should be skewed right,
though the skewness becomes less pronounced as the mean
increases
• Combine empirical tests with a good understanding of
how the data was generated

79
Example 4.33

• Suppose the number of deaths from typhoid fever over a

1-year period is Poisson distributed with parameter =4.6
1. What is the probability distribution of the number of
deaths over a 6-month period?

80
Example 4.33

oLet X=the number of deaths in 6 months

o=4.6 & t =1year, therefore λ = /t = 4.6 deaths per year
o6 month → t=0.5 year
onew = λ×(0.5)=4.6×(0.5)=2.3

81
Example 4.33
oLet X=the number of deaths in 6 months
o=4.6 & t =1year, therefore λ = /t = 4.6 deaths per year
o6 month → t=0.5 year
onew = λ×(0.5)=4.6×(0.5)=2.3
(2.3) k −2.3 2.3k −2.3
Pr( X = k ) = e = e , k = 0,1, 2,...
k! k!
• P(X=0)=.100
• P(X=1)=.231
• P(X=2)=0.265
• P(X=3)=.203
• P(X=4)=.117
• P(X=5)=.0543
• P(X>=6)=0.03
82
Example 4.33

• Suppose the number of deaths from typhoid fever over a

1-year period is Poisson distributed with parameter =4.6
1. What is the probability distribution of the number of
deaths over a 6-month period?
2. A 3-month period?
o3 month → t=0.25 year
onew = λ×(0.25)=4.6×(0.25)=1.15

83
Using R

oThere is a root name, for example, the root name for the
Poission distribution is pois
oThis root is prefixed by one of the letters
1. p for "probability", the cumulative distribution function
(c. d. f.)
2. q for "quantile", the inverse c. d. f.
3. d for "density", the density function (p. f. or p. d. f.)
4. r for "random", a random variable having the specified
distribution

84
85
Example 4.33
oLet X=the number of deaths in 6 months
o=4.6 & t =1year, therefore λ = /t = 4.6 deaths per year
o6 month → t=0.5 year
onew = λ×(0.5)=4.6×(0.5)=2.3
(2.3) k −2.3 2.3k −2.3
Pr( X = k ) = e = e , k = 0,1, 2,...
k! k!
dpois(0, 2.3)
• P(X=0)=.100
dpois(1, 2.3)
• P(X=1)=.231
• P(X=2)=0.265 dpois(2, 2.3)

• P(X=3)=.203 dpois(3, 2.3)

• P(X=4)=.117 dpois(4, 2.3)
• P(X=5)=.0543 dpois(5, 2.3)
• P(X>=6)=1-P(X=0)-…-P(X=5)=0.03 1-ppois(5, 2.3)

86
Binomial when n is large and p is very
small
o X~bin(n, p)
o μ(X) = E(X) = np
o σ2(X) = Var(X) = np(1-p)=npq
o If n is large and p is very small, 1-p = q ≈ 1
o Then np ≈ npq
o That is, E(X) ≈ Var(X)

87
Poisson approximation to the
Binomial distribution
• The binomial distribution with large n and small p can
be accurately approximated by a Poisson distribution
with parameter μ = np

88
• The Poisson distribution provides a good approximation for
the binomial distribution for n > 100 and p < 0.01:

𝑛 𝑘
lim 𝑃(𝑋 = 𝑘) = lim 𝑝 1 − 𝑝 𝑛−𝑘
𝑛→∞ 𝑛→∞ 𝑘
𝑛! 𝜇 𝑘 𝜇 𝑛−𝑘
= lim 1−
𝑛→∞ 𝑘! 𝑛 − 𝑘 ! 𝑛 𝑛
𝑛! 𝜇𝑘 𝜇 𝑛 𝜇 −𝑘
= lim 𝑘 1− 1−
𝑛→∞ 𝑛 𝑛 − 𝑘 ! 𝑘! 𝑛 𝑛
𝐹 ≈exp −𝜇 ≈1
𝜇𝑘
≈ exp −𝜇
𝑘!

89
 Example 4.39
• Suppose we are interested in the genetic susceptibility to
breast cancer
• From a large population study, we would expect that 1 in
1000 women aged 40-49 may develop a breast cancer over
one year of life
• We find in a sample that 4 out of 1000 women aged (40-49)
whose mothers have had breast cancer also develop cancer
over the next year of life
• How unusual is this event that there are 4 or more women
having breast cancer?

90
Using an exact Binomial distribution

• The exact Binomial probability could be computed by

letting n = 1000 and p = 0.001
Using the formula P( X = k ) = 1000  ( 0.001)k (1 − 0.001)1000−k
k 

• P(X ≥ 4) = 0.0189

91
Using a Poisson approximation

• We can also use the Poisson approximation with μ = np = 1 and

Thus P(X ≥ 4) = 1 − (0.3679 + 0.3679 + 0.1839 + 0.0613) = 0.0190

oThe Poisson approximation is indeed close
oThis event is unusual and suggests a genetic susceptibility to breast
cancer among the female offsprings of women who have had breast
cancer

92
Next class

• One-hour quiz from chapter 2~4

• Chapter 5: Continuous probability distributions

93