Probability Mass Function & Density

Function

By C K Nagpal
 Types of Random Variables
Discrete Random Variable
A random variable that takes finite number of values or
countable infinite number of possible values. For
example, random variable associated with events in the
toss of a coin or throw of dice, number of cars sold in a
day, number of persons visiting a cinema hall.
Continuous Random Variable
A random variable that takes infinite or non countable
number of values. For example, average score of a
subject in a class, a real number between 0 and 1.
Weight of a rice packet.
 Discrete Probability Distribution
Probability Distribution- A Probability
distribution for a random variable
specifies how the probabilities are
distributed over a random variable.
Given a random variable x, the
probability distribution is defined by a
probability function P(x) that provides
the value of probability for each value
of the random variable. Let there be a
car selling showroom, with sales
pattern for last 200 days as shown the
Table. The probability distribution for
car sales is shown in the last column.
 Mean or Expected Value
If an experiment is repeated
for a long period of time then
the average of the outcome is
most likely to approach a long
term expected or mean value.
The mean or expected value
of a discrete distribution is
given by the equation:
The expected value in this
case is 1.88. The expected
value is represented by E(x)
or μ.
 Probability Mass Function (PMF)
A PMF is a probability function which specifies the probability that any single
value of the discrete random variable will occur.
f(x)=P(X=x)
Here X is the set of all possible values.
x is a particular value of X.
P is the probability function.
f is the probability mass function.

For example, in the throw of a dice

f(2)=P(X=2)=1/6
f(4)=P(X=4)=1/6

For example, in the throw of a coin

f(Head)=P(X=Head)=1/2
Cumulative Mass Function(CMF)
A CMF is a probability function which specifies
the probability that observed value will not be
greater than x
F(x)=P(X ≤ x)
For example, in 4 throws of a coin chances of
getting upto 3 heads.
F(3)=P(X ≤ 3)=P(X=0)+P(X=1)+P(X=2)+P(X=3)
 Binomial Distribution
Most widely used discrete probability distribution.
Condition for applicability:
• Experiment involves a sequence of identical trials
• For each trial there are 2 possibilities: Success/failure, Head/tail etc.
• The trials are independent in nature
• If p is probability of the success then q=(1-p) is the probability of the
failure and these values remain constant throughout the experiment.
• Probability of r number of successes in n trials P(r):
 Probability Mass Function for Binomial
Distribution
Binomial experiment provides us with a
succinct formula for the Binomial
probability mass
function b(r;n,p) for r successes in n trials,
with p = the probability of success in each
trial.
Expected value for binomial probability
distribution=np
Variance for binomial probability
distribution= n.p.(1-p)=npq
Standard Deviation for binomial
probability distribution=
Q.1 The mean of a binomial distribution is 40
and the standard deviation is 6. Find n,p,q.

np=40 sqrt(npq)=6 npq=36

q=36/40=0.9 p=1-q=1-0.9=0.1
n=40/0.1=400
Q.2 Suppose half of the population in a town is consumer of rice.
100 investigators are appointed to check the truth. Each
investigators talks to 10 people. How many investigators will report
3 or less people as consumer of rice.
Q.3 Suppose 20% of the employees in a company suffer
from a particular occupational disease. Six workers are
checked at random. What is the probability that 4 or
more workers will be having the disease.
Q.4 Suppose 1 in 10 telephones is normally found busy. 6
telephones are randomly checked, what is the probability that
(a) exactly 3 are busy
(b)not more than 3 are busy
(c) more than 3 are busy.
 Poisson Distribution
Named after french mathematician: S Poisson
Characteristics:
1. Occurrence of events is independent. Occurrence of an
event in an interval of space is or time has no effect on its
second occurrence in the same interval or the different
interval.
2. Infinite number of occurrences of event are possible in a
given interval.
3. Probability of a single occurrence of the event in a given
interval is proportional to the length of time.
4. In any infinitesimally small interval the probability of
occurrence of two or more events is negligible.
Difference between Poisson & Binomial Distribution

1. Instead of discrete trials as in case of binomial

distribution, poisson distribution operates
continuously over some given amount of time,
area etc.
2. Instead of considering the sequence of
successes and failures, poisson distribution talks
about successes occuring at random points in a
given time or space context.
 Examples of Poisson Distribution
1. Demand of a product
2. Number of road accidents in a city
3. Footfall in a departmental store
4. Number of calls in a BPO
5. Number of faulty products in a production
system
 PMF or Poisson Distribution

Here m is called parameter of the distribution and is the

average number of occurrences ( successes). Here, x is the
number of occurrences of the event. As usual,

Mean and the variance for the Poisson Distribution is m

Q.1 In a production house, one in 400 items is defective. If the
items are packed in a box of 100. What is the probability that a
box will contain (i) no defective (ii) less than 2 defectives (iii) one
or more defectives (iv) more than 3 defectives
Q.2 A factory produces blades and packs in a packet of
10. If the probability of a defective blade is 0.2%. Find
the number of packets with 2 defective blades in
10000 packets.
Q.3 100 misprints are distributed randomly throughout the 100
pages of a book. What are the chances that a randomly selected
pages will have 3 misprints.

100 misprints on 100 pages.

Expected average number of misprints on a single
page=m=100/100 =1
 Negative Binomial Distribution(NBD)
Given the probability p for the success of an event, negative binomial
distribution talks about the likely number of trials needed for a given number of
successes. NBD is used to find the average waiting time.
Given a binomial distribution with success probability p, chances that kth
success will be achieved exactly in xth trial is:
(This is equivalent to saying that k-1 successes will be achieved in x-1 trials and
getting a success in xth trial).
= x-1Ck-1 pk-1 (1-p)(x-1)-(k-1) p = x-1Ck-1 pk (1-p)x-k

or P(x;k,p)=x-1Ck-1pk(1-p)x-k x=k,k+1,k+2,…….
=0 otherwise
(Here both x and k are integers and p lies between 0 and 1)
Above equation defines the PMF for NBD.
Mean for NBD=k/p
k (1  p )
p2
Variance=
Q.1 Given the success probability p=0.2 for an
Bernoullian trial find the expected number of
trials needed for 2 successes.

k
Mean for NBD is:
p

Here k=2 p=0.2

Likely or expected number of trials=2/0.2=10
Q.2 Given the success probability p=0.2 for an
Bernoullian trial find the probability that 2nd
successes will be achieved in 9th trial.
PMF for NBD is:
P(x;k,p)=x-1Ck-1pk(1-p)x-k x=k,k+1,k+2,…….
=0 otherwise
Here x=9,k=2 and p=0.2
P(x=9;k=2,p=0.2)=8C1(0.2)2(0.8)7 x=2,3,4,5,…….
=8 x 0.04 x 0.2097=0.067
Q.2 Given the success probability p=0.2 for an
Bernoullian trial find the probability that 2nd success
will be achieved in 2nd,3rd, 4th,5th , 6th ……….14th trial.

PMF for NBD is:

P(x;k,p)=x-1Ck-1pk(1-p)x-k
x=k,k+1,k+2,…….
=0 otherwise
Here x=2,3,4,5,6……14
k=2 and p=0.2
Q.4 A market research agency has observed that a call being made in
the afternoon has 0.4 probability of being answered. Find the
probability that
a) 10th answer comes in the 20th call
b) first answer comes in 4th call

a) P(x=20;k=10,p=0.4)=19C90.410 0.610 =0.05856

b) P(x=4;k=1,p=0.4)=3C00.41 0.63 =0.0864

 Multinomial Distribution (MD)
Multinomial Distribution is the generalization of Binomial distribution.
Here the random variable associated with the experiment has k > 2
possible discrete values.
Let E1,E2,E3,………Ek be the mutually exclusive possible outcomes of
the experiment in a single trial with the probability p1,p2,p3,………pk
respectively. Then, we have
p1,p2,p3,………pk ≥ 0.
p1+p2+p3,………+pk =1
Let xi denote the number of times event Ei occurs in n numbers of trials.
Then, we have
x1+x2+x3,………+xk =n
And the probability P for such a case is
Q.1 A company produces an item X. In the manufacturing
process, 6% of the items produced are undersized, 90% are
acceptable and 4% are oversized.
a)If a random sample of 100 is chosen then what are chances of
getting 2 undersized, 95 acceptable and 3 oversized.
a)If a random sample of 6 is chosen then what are chances of
getting 2 undersized, 3 acceptable and 1 oversized.
 Hypergeometric Distribution
A hypergeometric distribution arises when a sample is drawn
from a dichotomous population with replacement. A
dichotomous population is one, having two mutually exclusive
categories. For example red and blue balls in a bag, good and
defective items in a set.
Let there be n numbers of pieces of a manufactured product in a
bag with m numbers being defective. Then the probability of
finding k number of defectives in a sample size of s is:
m nm
Ck Cs  k
n
Cs
Q. A set of components are sent in a lot of 12. A sample of 3 is drawn from the lot and it is accepted
if none of the components is defective. What is the probability of a lot with 2 defectives being
accepted?
Distribution Associated With Continuous
Random Variable
Continuous Probability Function: A probability function
associated with continuous random variable.

Probability density function (PDF): An analog of the

probability mass function for discrete random variable but
is usually not used for specifying the probability for a
specific individual value.

Cumulative Density Function(CDF): An analog of the

cumulative mass density function for discrete random
variable. More useful than PDF and is used for defining the
probability that an observed value of the variable X will not
be greater than x.
CDF, F(x)=P(X ≤ x)
 Uniform Distribution
A continuous random variable X is
supposed to have a uniform
distribution with parameters a and
b, if its probability distribution
function is defined by
1
f(x)= b  a if a≤ x ≤ b

=0 otherwise

Mean=(b+a)/2
(b  a ) 2
12
Variance=
 Exponential Distribution
A continuous random variable X is
supposed to have a exponential
distribution with parameter λ if its
probability distribution function is
defined by

Mean= 1/ λ

Variance=