Probability Distributions

• Discrete Probability Distributions

• Binomial Distribution
• Poisson Distribution
• Negative Binomial Distribution
• Continuous Probability Distributions
• Uniform Distribution
• Exponential Distribution
• Normal Distribution

Gaurav Garg (IIM Lucknow)

• Example
• Consider a random experiment having only two outcomes :
• Success with probability p
• Failure with probability 1 − p
• Define a random variable as below:
• or
• pmf of can be written as
• or
• is said to have Bernoulli distribution.

Gaurav Garg (IIM Lucknow)

Bernoulli Distribution Discrete Distributions
• A discrete random variable is said to have a Bernoulli Distribution with
parameter , if its pmf is given by

• We write ~ Bernoulli () or
• Also,
• (show)
• (show)

Gaurav Garg (IIM Lucknow)

• Example:
• Consider the random experiment of tossing a fair die.
• Sample space: S = {1, 2, 3, 4, 5, 6}
• Let
• Success: getting “6”, i.e., {6}
• Failure: not getting “6”, i.e., {1, 2, 3, 4, 5}
• Clearly,
• P ( Success) = 1/6
• P(Failure) = 5/6
• If we repeat the tosses 3 times, i.e., we have 3 trials.
• The probability of success and failure will not change across trials.
• Trials are independent.

Gaurav Garg (IIM Lucknow)

• Let X: number of successes in 3 trials.
• X could take the values 0, 1, 2, or 3.

• P(X=0) = P(No Success in 3 trials) = P(FFF)

• = (5/6)(5/6)(5/6) = (5/6)3

• P(X=1) = P(1 success in 3 trials) = P (SFF or FSF or FFS)

• = P(SFF) + P(FSF) + P(FFS)
• = P(S)P(F)P(F) + P(F)P(S)P(F) + P(F)P(F)P(S)
• =(1/6)(5/6)(5/6) + (5/6)(1/6)(5/6) + (5/6)(5/6)(1/6)
• = 3. (1/6)1 (5/6)2

• P(X=2) = P(2 Successes in 3 trials) = P( SSF or SFS or FSS)

• = 3. (1/6)2 (5/6)1

• P(X=3) = P(3 Successes in 3 trials) = P(SSS) = (1/6)3

Gaurav Garg (IIM Lucknow)
• Probability distribution of is
x 0 1 2 3 Total
p(x) (5/6)3 3. (1/6)1 (5/6)2 3. (1/6)2 (5/6)1 (1/6)3 1

• When number of trials is large, the calculation becomes tedious.

• So, we define such probabilities through a general rule.
• This rule is applicable
• when a random experiment has a finite number of trials ()
• Each trial has only two complimentary outcomes – Success and Failure
• Probability of Success () remains constant across trials
• Each trial is independent of the others.
• Such trials are called as Bernoulli Trials.
• Here, the random variable = number of successes out of trials.
• Clearly,
• Probability of getting x successes in n trials is
Gaurav Garg (IIM Lucknow)
Binomial Distribution Discrete Distributions
• A discrete random variable is said to have a Binomial Distribution with
parameters and if its pmf is given by

• where
• We write

• We may show

Gaurav Garg (IIM Lucknow)

• Binomial distribution is a generalized case of Bernoulli distribution.
• Let are independent random variables and each
• Define
• Then,

•
•

Gaurav Garg (IIM Lucknow)

• Example:
• The production process of a company produces 5% defective items.
• If a random sample of 20 items is taken from the day’s production,
• what is the probability that
• there are no defective items in the sample;
• there is only one defective item in the sample;
• there is at least one defective item in the sample;
• there is at most one defective item in the sample.
• Example:
• The production process of a particular product produces 5% defective units.
• A customer has ordered a batch of 20 units.
• If the batch contains 3 or more defective units, the customer will reject the entire
batch, and cancel the order.
• What is the probability that the order will be canceled?
• What is the maximum permissible percentage (in integers, like 1%, 2%, etc.) of defectives in the
production process so that the probability of rejection is reduced to less than 1%?
Gaurav Garg (IIM Lucknow)
• Example:
• The management of a restaurant that operates on reservations only, knows from
experience that 15% of persons making table reservations will not show up.
• The restaurant accepts 25 reservations but has only 20 tables.
• What is the probability that all who show up will be accommodated?
• Example
• A passenger airline company has found from experience that 10% of the
customers who buy tickets for a flight do not show up for the journey.
• The company wishes to be at least 90% sure that a particular flight is at least 95%
full.
• How many minimum tickets should it sell if the capacity of the flight is 300?
• If the answer of the above question is k tickets, what is the probability that some
passenger(s) does(do) not get the seat?

Gaurav Garg (IIM Lucknow)

Poisson Distribution Discrete Distributions
• If

• When then

• A discrete random variable is said to have a Poisson Distribution with

parameters and if its pmf is given by
• where
• We write
• (show)

Gaurav Garg (IIM Lucknow)

• Thus, Poisson Distribution is a limiting case of Binomial Distribution.
• In practice, we approximate Binomial Distribution by Poisson Distribution
• when and ; or
• when and are almost equal.

• Practical Situations
• There are 100000 vehicles on the streets of Lucknow.
• On a given day, the probability of a vehicle meeting with an accident is 0.00005.
• accidents per day (accident rate)

• A Bank has 10000 customers.

• The probability of a customer arriving on a given day is 0.005.
• 0. Customers arrive at a rate of 50 per day (arrival rate).

• A machine produces 50000 parts per day, with 0.001 probability of a part being defective.
• 0 (average number of defective parts produced in a day)
Gaurav Garg (IIM Lucknow)
• The Poisson distribution and Binomial Distribution have some similarities
and differences.
• Binomial Distribution describes a distribution of two possible outcomes – Success
and Failure.
• Poisson Distribution focuses only on the number of discrete occurrences over some
interval.
• A Poisson experiment does not have a given number of trials as a binomial
experiment does.
• The Poisson describes the occurrence of rare events. E.g., no. of aircraft accidents.
• If the occurrences are too frequent, the time interval can be reduced enough so that the
occurrences are rare.
• Other Applications:
• No. of deaths from a disease (not in the form of an epidemic);
• No. of suicides reported in a particular city;
• No. of defective items in a packing manufactured by a good concern;
• No. of printing mistakes on each page of a book.
Gaurav Garg (IIM Lucknow)
• Example:
• In a book of 520 pages, 390 typographical errors occur.
• Assuming Poisson probability law for the number of errors per page, find the
probability that a random sample of 5 pages will contain no error.
• Ans: e – 3.75
• Example:
• Six coins are tossed 6400 times.
• Using the Poisson Distribution, find the approximate probability of getting six
heads 150 times.
• Ans:

Gaurav Garg (IIM Lucknow)

• Example:
• A manufacturer, who produces medicine bottles, finds that 0.1% of the bottles are
defective.
• The bottles are packed in boxes containing 500 bottles.
• A drug manufacturer buys 100 boxes from the manufacturer of bottles.
• Using Poisson distribution, find how many boxes are expected to contain:
• no defective;
• at most two defectives.
• Example:
• A toll-free phone number is available from 9:00 AM to 9:00 PM for the customers to
register complaints about a product purchased from a company.
• Past history indicates that an average of 0.4 calls are received per minute.
• What is the probability that during a 1-minute period
• Zero phone calls will be received?
• 3 or more phone calls will be received?
• At most, how many phone calls will be received in a 1-minute period 99.99% of the time?
Gaurav Garg (IIM Lucknow)
• Additive Property of Poisson Distribution:
• If and are independent random variables.

• Then Poisson
• Example:
• Suppose that the number of calls at a customer care unit between 10:00 AM and
11:00 AM is a random variable X.
• X~ Poisson (2).
• Similarly, the calls arriving between 11:00 AM and 12:00 Noon is Y.
• Y ~ Poisson (6).
• X and Y are independent
• What is the probability that more than 5 calls come between 10:00 AM and 12:00
Noon?
• Ans: 0.8088
Gaurav Garg (IIM Lucknow)
Negative Binomial Distribution Discrete Distributions
• If we toss a coin a fixed number of times (say times).
• Number of heads out of tosses is a random variable following a binomial distribution.
• If we continue tossing the coin until we get a fixed number of heads (say r heads).
• Number of tosses needed to get heads is a random variable.
• This random variable is said to have a Negative Binomial Distribution.
• In general,
• we have a random experiment having two complimentary outcomes – Success and Failure
• Probability of Success (𝑝) remains constant across trials
• Trials of the experiment are independent of the one another.
• We repeat the trials till we get Successes.
• The number of trials
• Then,

•
•
•
Gaurav Garg (IIM Lucknow)
• A discrete random variable X is said to have a Negative Binomial
Distribution if its pmf is given by

• We write
• (!!!)
• (!!!)
• A particular case
• When is said to have a Geometric distribution with parameter .
• Negative Binomial Distribution may also be defined as the distribution of
=number of failures

Gaurav Garg (IIM Lucknow)

• Example
• A sculptor is making 3 exhibits for an art gallery.
• There is a probability of 0.75 that every piece of wood she carves into will
be good enough to be part of the exhibit.
• What is the probability that she uses 4 pieces of wood or less?

Gaurav Garg (IIM Lucknow)

Hypergeometric Distribution

Sample
Population X Red Balls + (n-X) White Balls
A Red Balls + (N – A) White Balls

• X is a discrete random variable.

• X ≤ A and X ≤ n
• So, X ≤ min {n, A}
• n-X ≤ N-A and n-X ≤ n orX ≥ n-N+A and X ≥ 0
• So, X ≥ max (0, n-N+A) Gaurav Garg (IIM Lucknow)
• Let us assume a bag having red and white balls.
• In the bag there are A red balls and rest white balls.
• A sample of size is selected from the bag one by one without replacing
them back in the bag.
• Let be the number of red balls selected in the sample.
• X is a discrete random variable.

• pmf of is

• is said to follow Hypergeometric Distribution with parameters and .

Gaurav Garg (IIM Lucknow)

• Example:
• In a shipment of 15 sets of golf clubs, 3 are left-handed. If 4 sets of golf clubs are
selected (without replacement), what is the probability that
• Exactly 1 is left-handed.
• At least 1 is left-handed.
• No more than 2 are left-handed.

Gaurav Garg (IIM Lucknow)

Uniform Distribution Continuous Distributions
• X is a continuous random variable.
• X takes the values in the interval (a,b)
• All values of X are equally likely. (Uniform)
• Probability Density is the same everywhere.
• A continuous random variable X is said to follow Uniform Distribution
with parameters a and b (−∞ < a < b < ∞), if its pdf is

• We write X~U(a, b)
• (show)
• Var (show)

Gaurav Garg (IIM Lucknow)

• cdf of

• Uniform Distribution is also called as rectangular distribution.

1/(b-a)
a t b

P(X ≤ t)
Gaurav Garg (IIM Lucknow)
• Example
• A company manufactures field rifles for the Defense forces.
• One of the production lines is producing firing pins with length varying from 0.85 inch
to 1.05 inch with equal probability.
• Firing pins with length less than 0.90 inch or greater than 1.00 inch do not meet the
specifications and must be discarded.
• Find the fraction of the pins that must be discarded.

• Required Fraction =
• Example
• Department of Transportation (DOT) invites the bids for the construction of a road.
• The DOT estimate of the cost of the job is .
• The qualifying bid can take any value in the interval with equal probabilities.
• What fraction of qualifying bids on road construction contracts are less than the DOT
estimate?
• 𝑋 ~ 𝑈 (2𝑑/5, 2𝑑)
• Ans = 𝑃(𝑋< 𝑑)×100 %= 37.5%
Gaurav Garg (IIM Lucknow)
Exponential Distribution Continuous Distributions
• Exponential distribution is used widely to model
• Waiting time in a bank’s queue;
• Life time of some component or human being; etc.

• A continuous random variable is said to have Exponential Distribution

with mean if its probability density function is given by

• We write

• Var

• is usually the average number of occurrences (or arrivals) per unit time.
Gaurav Garg (IIM Lucknow)
• Example
• A processing machine at Wiley Inc. breaks down on an average of once in four
weeks.
• What is the probability that the next breakdown will not occur for at least six
weeks after the previous breakdown?
• Assume that the time between breakdowns has an exponential distribution.

• Example
• A teller at New York Industrial Bank serves, on average, 30 customers per hour.
• Assume that the service time for a customer has an exponential probability
distribution.
• What is the probability that the next customer will take two to four minutes to be
served.
Gaurav Garg (IIM Lucknow)
• Memoryless Property of exponential distribution
• has exponential distribution with mean
• for any

• Exponential distribution is used to model customer arrivals in queuing systems

• Probability that a customer will arrive in next 5 minutes

• Suppose that no customer arrives in the 5 minutes.

• Given this, the probability that a customer will arrive in next 5 minutes

• After 10 minutes:

• For exponential distribution, it is the same as , no matter how long you have waited.
Gaurav Garg (IIM Lucknow)
• Relationship between Poisson and Exponential Distributions

• Let be the number of events occurring during the time interval .

• is the number of occurrences of the event in the unit time interval,

• If
• Then, the time-gap between any two successive occurrences follows Exp().

Gaurav Garg (IIM Lucknow)

Normal Distribution Continuous Distributions
• Normal is the most widely used probability distribution.
• Normal distribution is used to give approximated probabilities underlying
many other distribution.
• A continuous r. v. X is said to have a Normal Distribution with parameters
μ and σ, if its pdf is

• We write X ~ N(μ, σ ).

• Var

Gaurav Garg (IIM Lucknow)

• The bell-shaped curve depicts the normal distribution.
• For the normal distribution, mean=median=mode.
• Normal distribution has a wide application in practice.
• Data on many characteristics have a bell shape, provided data size is large.
• Probabilities of many discrete as well as continuous distributions can be
approximated using normal distributions in large sample cases.
• Central limit theorem:
• Consider taking a sample of size n from the population that may be non-normal.
• The distribution of the sample mean is normal when n is large enough.

μ
Gaurav Garg (IIM Lucknow)
 = 10  = 10  = 10

=5  = 10  = 15

=5

 = 10
 = 15

=0

=5
 = 15  = 10

=5  = 10  = 15
Gaurav Garg (IIM Lucknow)
• In a continuous distribution
• Probability is given by the area below the probability curve.

Gaurav Garg (IIM Lucknow)

Normal Probability Curve is symmetric about mean, i.e.,
 μ = 100
σ = 30

Area =0.341345

100 130
σ = 30

μ = 60
σ = 35 Area = 0.341345

60 95
σ = 35
• For a N(,  ) distribution
• approximately 68% of all the values lie in (-, +)
• approximately 95.5% of all the values lie in (-2, +2)
• approximately 99.7% of all the values lie in (-3, +3)

16% of 16% of 2.25% of 2.25% of

area 68% area area 95.5% area
of area of area

σ σ 2σ 2σ

.15% of .15% of
area 99.7% area
of area

3σ 3σ
• Linearity Property of Normal Distribution

• are constants
• s a linear function of
• Then,
• Standard Normal Distribution
• Choose
• Then,
• Thus, we can transform any distribution to distribution
• is called standard normal distribution
• is called standard normal variable
 X~N(10, σ)

(X+5)~N(15, σ)
(X-5)~N(5, σ)

=5  = 10  = 15

X/2~N(0, 5)

X~N(0, 10)
2X~N(0, 20)

=0
 z value corresponding to a

N(0,1) N(,)

Za= (a-)/ a
• Computing Normal Probabilities

• Alternatively,
• Values of depend on .
• Statistical Table of values is available. Here
• Therefore,
•

a b
F(a) F(b)
• Example
• Daily cement production of a plant follows normal distribution with mean of 50
tons and standard deviation of 3.5 tons.
• On how many days in a year the production is expected to exceed 55 tons.

• Clearly, Daily cement production:

• Ans = 365 x 0.0764 = 27.9443 days

Gaurav Garg (IIM Lucknow)

• Example
• A group of achievement scores are normally distributed with a mean of 76 and a
standard deviation of 4.
• If one score is randomly selected what is the probability that it is at least 80.

 4

76 80

• Ans = 0.1587
• Example:
• Time required to finish an exam is known to be normally distributed with a mean
of 60 Min. and a Std Dev. of 12 minutes.
• How much time should be allowed in order for 90% of the students to finish?

.9  12

60 x

Gaurav Garg (IIM Lucknow)

• Example
• The marks on a statistics test are normally distributed with a mean of 62 and a
variance of 225.
• The instructor wishes to assign ‘B’ or higher to the top 30% of the students in the
class.
• Obtain minimum marks required to get a ‘B’ or higher?
• Example
• A bank manager has learnt that the length of time the customers have to wait for
being attended by the teller is normally distributed with mean time of 5 minutes
and standard deviation of 0.8 minutes.
• Find the probabilities that a customer has to wait:
• for less than 6 minutes
• for more than 3.5 minutes
• between 3.4 and 6.2 minutes

Gaurav Garg (IIM Lucknow)

• Example
• One of the side effects of flooding a lake is that mercury leached from the soil,
enters the food chain, and eventually contaminates the fish.
• The concentration in fish will vary among individual fish because of differences in
eating patterns, movements around the lake, etc.
• Suppose that the concentrations of mercury in individual fish follows an
approximate normal distribution with a mean of 0.25 ppm and a standard
deviation of 0.08 ppm.
• Fish are safe to eat if the mercury level is below 0.30 ppm.
• What proportion of fish are safe to eat?
• Normal Approximation of Binomial
• Let X~B(n, 1/4), following are the probability plots of X for various values of n

n=4 n=10

n=25
n=20
Gaurav Garg (IIM Lucknow)
• Let X ~ B(n, p),

• Then for sufficiently large

• , or

• In practice, the result is valid when

• , or

• Correction for continuity:

• While approximating a discrete distribution by a continuous one.
• Subtract 0.5 from the lower limit.
• Add 0.5 to the upper limit.
• is approximated by
• is approximated by

Gaurav Garg (IIM Lucknow)

• Example
• According to an estimate, 50% of the people in America have at least one credit
card.
• If a random sample of 30 persons is taken,
• what is the probability that 19 of them have at least one credit card?

• n=30, p=0.5, q=0.5.

• X=number of persons in the sample of 30 having at least one credit card.
• Then X~B(30,0.5)
• So, P(X=19)= 30C19 (0.5)19(0.5)30-19 =0.0509
• np=15, nq=15, so we can approximate this probability by normal distribution.

Gaurav Garg (IIM Lucknow)

Gaurav Garg (IIM Lucknow)
• Example
• A fair coin is tossed 40 times
• Let X be the number of times it lands heads.
• Find P(X=20).
• Use normal approximation and compare it to the exact solution.

• Using Binomial Distribution, Exact Solution is

• P(X=20)=0.1254

• Using Normal Approximation

• P(X=20) = P(19.5<X<20.5)
• = 0.1272

Gaurav Garg (IIM Lucknow)

• Normal Approximation of Poisson
• Let

• For sufficiently large , we have

• , or

• Practically when > 10, Normal is a reasonable approximation of Poisson.

• Example
• If X follows Poisson distribution with = 25, find the probability of X = 20 using normal
approximation.

• P(X=20) = P(19.5 < X < 20.5) = P(-1.1 < Z < -0.9) = 0.1841 – 0.1357 = 0.0484

• P(X = 20) = 0.0519 using Poisson Formula

Gaurav Garg (IIM Lucknow)

• Examining Normality of a given Data set
• Constructing Q-Q plot
• A plot of data scores against the expected normal scores
• The expected normal scores are calculated by taking the z-scores of I/(n+1), where I is the rank
in increasing order.
• Curvature of the points indicates departures of normality.
• Constructing Histogram
• A frequency plot having columns with the heights representing frequencies
• Should have a bell like shape for normality.
• Constructing Box and Whisker Plot
• Provides a representation of
• Smallest and largest observations
• Median, First and third quartile
• Comparing theoretical properties
• Mean = Median = Mode
• Range = 6 σ approximately
• Interquartile range = 1.33σ approximately
Gaurav Garg (IIM Lucknow)
Probability Distributions

Exponential ()

n  ∞, p 0
Poisson ()
Bernoulli Trial Binomial (n, p)
n ≥ 20, p ≤ .05
 large
n large  > 10
np > 5, nq > 5
Normal (, )

??

Uniform (a, b)

Gaurav Garg (IIM Lucknow)