MATHEMATICS FOR

ELECTRONICS ENGINEERS
CONTINUOUS RANDOM VARIABLES

First, A Recap
RANDOM VARIABLE
Is a Function

• A random variable X is a function that maps Ω to ℝ, i. e.,

X : Ω →ℝ such that 𝜔: 𝑋(𝜔) ≤ 𝑥 ∈ E, ∀ 𝑥 ∈ ℝ
• 𝜔: 𝑋(𝜔) ≤ 𝑥 ∈ E ⇒ we can assign probability value to it
• 𝑃 𝜔: 𝑋(𝜔) ≤ 𝑥 is abbreviated as P(𝑋 ≤ 𝑥), 𝑥 is the variable in control
• Random variables (RV) can be classified as discrete RV and continuous RV,
depending on whether Ω is a discrete set or a continuous set
DISCRETE AND CONTINUOUS SETS
The Concept of Infinity Is Way Beyond Your Intuition!

• A set is said to be discrete, if its cardinality is either finite or countable

Examples: Tossing of a coin: Ω={H,T} - Bernoulli
Picking a random person on earth: Ω={1,2,…,P} - Binomial
Number of possible phone calls during your life: Ω={1,2,…} – Poisson
• A set is said to be continuous, if its cardinality is uncountable
Examples: Waiting time for a taxi
Amount of rainwater harvested
Delays caused at an office
Radioactive half-life
Communication: Noise and channels
CONTINUOUS RANDOM VARIABLES

Defined over Continuous Sample Space Ω

Ω is said to be continuous if its cardinality is uncountably infinite
CONTINUOUS RANDOM VARIABLES

Associated Functions
CUMULATIVE DISTRIBUTION FUNCTION

Accumulates Probability Values

Carefully notice the difference between capital X and little x
CUMULATIVE DISTRIBUTION FUNCTION (CDF)
Definition and Properties

The CDF of a continuous random variable X is defined as

𝐹𝑋 𝑥 ≜ 𝑃 𝑋 ≤ 𝑥 , ∀ 𝑥 ∈ ℝ.
Properties:
1. CDF is monotonically non-decreasing function in 𝑥, i. e.,
𝐹𝑋 𝑥2 ≥ 𝐹𝑋 𝑥1 , if 𝑥2 > 𝑥1, ∀ 𝑥1, 𝑥2 ∈ ℝ

2. lim 𝐹𝑋 𝑥 = 0 𝑎𝑛𝑑 lim 𝐹𝑋 𝑥 = 1.

𝑥→−∞ 𝑥→∞
3. 𝐹𝑋 𝑥 is right continuous in 𝑥, for instance

1
lim 𝐹𝑋 𝑥+ = 𝐹𝑋 𝑥 .
𝑁→∞ 𝑁
4. 𝑃 𝑥1 < 𝑋 ≤ 𝑥2 = 𝐹𝑋 𝑥2 − 𝐹𝑋(𝑥1)
 CUMULATIVE DISTRIBUTION FUNCTION (CDF)
Problems
Which of the following functions could be CDF for some random variable?
1 1
1. 𝐹𝑋 𝑥 = + tan −1 𝑥
2 𝜋

2. 𝐹𝑋 𝑥 = 1 − 𝑒−𝑥 𝑢 𝑥
2
3. 𝐹𝑋 𝑥 = 𝑒−𝑥
4. 𝐹𝑋 𝑥 = 𝑥2 𝑢 𝑥
Solution:
• To determine this, we need to check that the function starts at 0 when 𝑥 = −∞, ends at 1 when
𝑥 = ∞, and is monotonic increasing in between. CDF
• The first two functions satisfy these properties and thus are valid CDFs,
• while the last two do not. The function in (3) is decreasing for positive values of x, while the
function in (4) takes on values greater than 1 and 𝐹𝑋(∞) = 1.
CUMULATIVE DISTRIBUTION FUNCTION (CDF)
Problems
CUMULATIVE DISTRIBUTION FUNCTION (CDF)
Problems

Suppose a random variable has a CDF given by

𝐹𝑋 𝑥 = 1 − 𝑒−𝑥 𝑢 𝑥
a) 𝑃 𝑋 > 5 ,
b) 𝑃 𝑋 < 5 ,
c) 𝑃(3 < 𝑋 < 7),
Solution:

a) 𝑃 𝑋 > 5 = 1 − 𝑃 𝑋 ≤ 5 = 1 − 𝐹𝑋 5 = 𝑒−5
b) 𝑃 𝑋 < 5 = 𝑃 𝑋 ≤ 5 = 1 − 𝑒−5 (Since 𝑃 𝑋 = 5 = 0)
c) 𝑃 3 < 𝑋 < 7 = 𝑃 3 ≤ 𝑋 ≤ 7 = 𝐹𝑋 7 − 𝐹𝑋 3 = 𝑒−3 − 𝑒−7
PROBABILITY DENSITY FUNCTION

Probability “Mass Density” Values

Carefully notice the difference between capital X and little x
PROBABILITY DENSITY FUNCTION (PDF)
Definition

• Note that the definition of a PMF of a continuous random variable X is

meaningless. That is, for a continuous random variable
𝑝𝑋 𝑥 ≜ 𝑃 𝑋 = 𝑥 , ∀𝑥 ∈ℝ
does not make sense!

• The concept of mass on a point in the continuum does not make sense
• Hence, we need to define a certain mass density-type meaning to evaluate
the probability values, over a given interval 𝑎, 𝑏 𝜖 ℝ
• There is no definition of PDF, but we call a function as a valid PDF, if it
satisfies some properties
PROBABILITY DENSITY FUNCTION (PDF)
Definition

• Probability Density function of a Random Variable at a point

𝑋 evaluated at a point 𝑥 is

𝑃(𝑥 ≤ 𝑋 < 𝑥 + 𝜖)
𝑓𝑋 𝑥 = lim
𝜖 →0 𝜖
• The probability density function is the probability that the random
variable 𝑋 lies in an infinitesimal interval about the point 𝑋 = 𝑥,
normalized by the length of the interval.
PROBABILITY MASS FUNCTION (PMF)
Recap

Recall that the PMF for a discrete random variable has the following properties

1. 𝑝𝑋 𝑥 ≥ 0, ∀ 𝑥 ∈ ℤ.

2. σ ∞
𝑥=−∞ 𝑝𝑋 𝑥 = 1

3. σ 𝑥𝑎=−∞ 𝑝𝑋 𝑎 = 𝐹𝑋 (𝑥)
PROBABILITY DENSITY FUNCTION (PDF)
Properties
PROBABILITY DENSITY FUNCTION (PDF)
Properties

Following the property 3, it holds that

𝑃(𝑥 ≤ 𝑋 < 𝑥 + 𝜖)
𝑓𝑋 𝑥 = lim
𝜖 →0 𝜖

𝐹𝑋 𝑥 + 𝜖 − 𝐹𝑋(𝑥)
= lim
𝜖 →0 𝜖

𝑑𝐹𝑋 𝑥
𝑓𝑋 (𝑥) =
𝑑𝑥
PROBABILITY DENSITY FUNCTION (PDF)
Problems

Find the constant 𝑏 > 0 so that the function

𝑒3𝑥
𝑓𝑋 𝑥 = 4 ;0 ≤ 𝑥 ≤ 𝑏
0 ; 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
is a valid probability density.
PROBABILITY DENSITY FUNCTION (PDF)
Problems
PROBABILITY DENSITY FUNCTION (PDF)
Problems
PROBABILITY DENSITY FUNCTION (PDF)
Problems
Example 1
Example 1
TO UNDERSTAND A RANDOM VARIABLE

A Continuous Random Variable is Completely

Characterized By its CDF or PDF!
RANDOM VARIABLES

Well-Known Continuous Random Variables

RANDOM VARIABLES
List of Well-Known Continuous Random Variables

A continuous random variable is completely characterized by

either its CDF or PDF

1. Uniform random variable

2. Gaussian random variable
3. Exponential random variable
4. Rayleigh random variable
5. Chi-squared random variable
6. There are several others, but we consider only the above
GAUSSIAN / NORMAL RANDOM VARIABLE
Notation: 𝑿~𝓝(𝝁, 𝝈𝟐)

• Most commonly used continuous random variable

• Takes values on the entire real line 𝑋 𝜖 ℝ
• The PDF looks like (famous bell-shaped curve)
𝐹𝑋(𝑥)
𝑓𝑋 (𝑥)
1

0.5

𝑥 𝑥
𝜇 0
• The PDF is given by
1 − 𝑥−𝜇 2
𝑓𝑋 𝑥 = 𝑒 2𝜎 2
2𝜋𝜎2
GAUSSIAN / NORMAL RANDOM VARIABLE
𝑿~𝓝 𝝁, 𝝈𝟐 : CDF
GAUSSIAN / NORMAL RANDOM VARIABLE
𝑿~𝓝 𝝁, 𝝈𝟐 : CDF
𝑿~𝓝 𝝁, 𝝈𝟐 : Example 1

Find the probability of the event 𝑋 ≤ 5.5 for a gaussian

random variable having 𝑎𝑋 = 3 𝑎𝑛𝑑 𝜎𝑋 = 2.
Solution:
wkt,
𝑥 − 𝑎𝑋
𝐹𝑋 𝑥 = 𝐹
𝜎𝑋

5.5 − 3
𝐹𝑋 5.5 = 𝐹
2

= 𝐹(1.25)
= 0.8944
GAUSSIAN / NORMAL RANDOM VARIABLE
𝑿~𝓝 𝝁, 𝝈𝟐 : Example 2

Assume that the height of clouds above the ground at some location is
a gaussian random variable X with 𝑎𝑋 = 1830 𝑚 𝑎𝑛𝑑 𝜎𝑋 = 460 𝑚.
Find the probability that clouds will be higher than 2750 m.
Solution:
𝑃 𝑋 > 2750 = 1 − 𝑃 𝑋 ≤ 2750
= 1 − 𝐹𝑋 2750
2750−1830
=1−𝐹 , (𝑆𝑖𝑛𝑐𝑒 , 𝐹𝑋 𝑥 = 𝐹(𝑥−𝑎𝑋))
460 𝜎𝑋

= 1 − 𝐹 2.0 = 1 − 0.9772 = 0.0228

𝑿~𝓝 𝝁, 𝝈𝟐 : Example 3

b)𝑃 𝑋 > 42 = 1 − 𝑃 𝑋 ≤ 42

= 1 − 𝐹𝑋 42

42−30 𝑥−𝑎 𝑋
=1−𝐹 , (𝑆𝑖𝑛𝑐𝑒 , 𝐹𝑋 𝑥 = 𝐹( ))
5 𝜎𝑋

= 1 − 𝐹 2.4 = 1 − 0.9918 = 0.0082

Example 3
If 𝑋 is distributed over (0,10), calculate the following
a) 𝑃 𝑋 > 8
b) 𝑃 𝑋 < 3
c) 𝑃 3 < 𝑋 < 8

Solution: Using CDF

0; 𝑥<0
𝑥 ; 0 < 𝑥 < 10
𝐹𝑋 𝑥 =
10
1; 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
a)
𝑃 𝑋 >8 =1−𝑃 𝑋 ≤8
= 1 − 𝐹𝑋 8
8
=1−
10
2
=
Example 3

b)
𝑃 𝑋 < 3 = 𝑃 𝑋 ≤ 3 = 𝐹𝑋 3
3
=
10
c)
𝑃 3 < 𝑥 < 8 = 𝐹𝑋 8 − 𝐹𝑋 3
8 3
= −
10 10
1
=
2
GAUSSIAN / NORMAL RANDOM VARIABLE
𝑿~𝓝(𝝁, 𝝈𝟐): Examples and Applications

• Measurement errors in physical experiments

• Confidence level calculation in diabetes test results
• Scores in a common entrance test
• Additive noise in communication systems
• Information theory
• Central limit theorem
RANDOM VARIABLES

Well-Known Continuous Random Variables

MOMENTS AND EXPECTATION

Non-Central Moments
Central Moments
Expectation
Variance
 Operations on a Single Random Variable

• The PDF and CDF provides a complete statistical description of the random variable.
• In theory, we can determine anything we might want to know about the random
variable.
• In many cases, it is of interest to distill this information down to a few parameters that
describe some of the important features of the random variable.
• Gaussian random variable is described by two parameters, which were referred to as
the mean and variance.
• We will look at these parameters as well as several others that describe various
characteristics of random variables.
• These parameters can be viewed as the results of performing various operations on a
random variable.
EXPECTATION

First Non-Central Moment

“Mean / Average”
 EXPECTATION
Definitions/Meaning

• Expected value of a discrete random variable is simply a weighted average of

the values that the random variable can take on, weighted by the probability mass
of each value.

𝔼 𝑋 stands for expectation of 𝑿

𝔼 𝑋 represents the “mean/average” of 𝑋.
Expectation
Problems

A discrete random variable X has possible values 𝑥𝑖 = 𝑖2, 𝑖 = 1,2,3,4,5 which

occurs with the probabilities 0.4, 0.25, 0.15, 0.1 and 0.1 respectively. Find the
mean value of X.
 EXPECTATION
Examples

• Determine the mean value of the continuous RV for which

EXPECTATION
Examples
EXPECTATION
Properties

• Lemma 1: If 𝑋 is a random variable and 𝑔 . is a function, then

𝑔(𝑋) is also a random variable
• Theorem 1: Expectation is a linear operator. That is, if 𝑎 and 𝑏 are
real numbers, then for a random variable 𝑋,
𝔼 𝑎𝑋 + 𝑏 = 𝑎 𝔼 𝑋 + 𝑏
Let us prove this result for a discrete random variable.
• Theorem 2: For a given random variable 𝑋 and a function 𝑔(. ),
the expectation of 𝑔(𝑋) is given by
• Expectation of a constant E(a)=a E(5)=5
Expectation of a constant is always a constant
• Linear combination/Linear Property
𝔼(𝑎𝑋 ± 𝑏Y)= 𝔼(𝑎 𝑋) ± 𝔼(𝑏 Y) =𝑎 𝔼(𝑋) ± 𝑏 𝔼(Y)
EXPECTATION
Other Properties and Notation

• Obviously – given the interpretation of mean / average,

expectation of a random variable is a real number
• Nothing is random about the expectation

• Notation: We will use 𝔼 𝑋 and 𝔼𝑋 interchangeably

• Notations should be clear from the context
NON-CENTRAL MOMENTS

Raw Moments
NON-CENTRAL MOMENTS
Definitions
CENTRAL MOMENTS

Moments Around Mean

MOMENTS AND EXPECTATION

Central Moments
Expectation
Variance

Department of Electronics and Communication Engineering

CENTRAL MOMENTS

Moments Around Mean

Department of Electronics and Communication Engineering

Definition Moments Around Mean

CENTRAL MOMENTS
• The nth central moment of a discrete random variable is
defined as

Very carefully note the difference between 𝑥 and 𝑋

CENTRAL MOMENTS
Definitions
VARIANCE

Second Central Moment

On An “Average”, Farness From The “Average”

Rajini M.
Department of Electronics and Communication Engineering
VARIANCE
Definitions/Meaning
VARIANCE
Properties

• Property 1: Variance of a random variable 𝑋 is non-negative

• Property 2: Variance of a constant is zero
• Property 3: Variance is not a linear operator. However, variance is
additive. For two random variables 𝑋 and 𝑌,
𝑣𝑎𝑟 𝑋 + 𝑌 = 𝑣𝑎𝑟 𝑋 + 𝑣𝑎𝑟(𝑌)
• Property 4: For a random variable 𝑋
𝑣𝑎𝑟 𝑋 = 𝔼 𝑋2 − 𝔼𝑋 2
• Property 5: For a random variable 𝑋
𝑣𝑎𝑟(𝛼𝑋 + 𝛽) = 𝛼 2 𝑣𝑎𝑟(𝑋)
 Problems -Non central moments

A random variable has the density

Find the following moments i) 𝑚0, ii) 𝑚1, iii) 𝑚2 and iv) 𝜇2
Problems

• Determine the expectation and variance of the following discrete random

variable which has the PMF
• (i) Bernoulli RV: 𝑃 𝑋 = 1 = 𝑝
𝑃 𝑋 =0 =1−𝑝
 Problems

• Determine the expectation and variance of the following discrete random

variable which has the PMF
• Binomial RV: 𝑃 𝑋 = 𝑘 = 𝑘𝑛 𝑝𝑘 1 − 𝑝 𝑛−𝑘 , 𝑘 = 0,1, … , 𝑛
−𝜆 𝑘
• Poisson RV: 𝑃 𝑋 = 𝑘 = 𝑒 𝜆
𝑘 = 0,1, …
𝑘!
Expectation and Variance of Binomial RV
Variance of Binomial RV
Expectation and Variance of Poisson RV
 Expectation of Gaussion Random Variable

Determine the expectation and variance of Gaussian Random variable with PDF
2
1 − 𝑥−𝑎2
𝑓𝑋 𝑥 = 𝑒 2𝜎
2𝜋𝜎
Expectation of Gaussian RV
Variance of Gaussian Random Variable
Variance of Gaussian Random Variable
MOMENTS-RELATED FUNCTIONS

Moment Generating Functions

Characteristic Functions
MOMENT GENERATING FUNCTIONS

Generate “All” Moments In One-Shot

EXPECTATION
Recall Theorem 2 Among Its Properties
MOMENT GENERATING FUNCTION (MGF)
Definition
MOMENT GENERATING FUNCTION (MGF)
Taylor Series Expansion
MOMENT GENERATING FUNCTION (MGF)
Derivatives and Moments
MOMENT GENERATING FUNCTION (MGF)
Relationship with Laplace Transform
MOMENT GENERATING FUNCTION (MGF)
𝑩𝒊𝒏(𝒏, 𝒑)
 MOMENT GENERATING FUNCTION (MGF)
𝑩𝒊𝒏(𝒏, 𝒑)

𝑀𝑋(𝑡) = 𝑝𝑒𝑡 + 1 − 𝑝 𝑛

Differentiating,

𝑀𝑋′ 𝑡 = 𝑛 𝑝𝑒𝑡 + 1 − 𝑝 𝑛−1 𝑝𝑒𝑡

𝐸 𝑥 = 𝑀𝑋′ 0 = 𝑛𝑝

𝑀𝑋′′ 𝑡 = 𝑛 𝑛 − 1 𝑝𝑒𝑡 + 1 − 𝑝 𝑛−2 𝑝𝑒𝑡 2 + 𝑛 𝑝𝑒𝑡 + 1 − 𝑝 𝑛−1 𝑝𝑒𝑡

𝑀𝑋′′ 0 = 𝐸 𝑋2 = 𝑛 𝑛 − 1 𝑝2 + 𝑛𝑝

𝑉𝑎𝑟 𝑋 = 𝐸 𝑋2 − 𝐸 𝑋 2 = 𝑛𝑝(1 − 𝑝)
MOMENT GENERATING FUNCTION (MGF)
𝑷𝒐𝒊(𝝀)
MOMENT GENERATING FUNCTION (MGF)
𝑷𝒐𝒊(𝝀)

𝑀𝑋 𝑡 = exp 𝜆 𝑒 𝑡 − 1

Differentiating,
𝑀𝑋′ 𝑡 = 𝜆𝑒𝑡 exp 𝜆 𝑒 𝑡 − 1
𝑀𝑋′′ 𝑡 = 𝜆𝑒𝑡 2 exp 𝜆 𝑒𝑡 − 1 + 𝜆𝑒𝑡 exp 𝜆 𝑒 𝑡 − 1

We have,
𝐸 𝑋 = 𝑀𝑋′ 0 = 𝜆
𝐸 𝑋2 = 𝑀𝑋′′ 0 = 𝜆2 + 𝜆
𝑉𝑎𝑟 𝑋 = 𝜆2 + 𝜆 − 𝜆2 = 𝜆
MOMENT GENERATING FUNCTION (MGF)
𝓝(𝟎, 𝟏)
MOMENT GENERATING FUNCTION (MGF)
𝓝(𝟎, 𝟏)
 MOMENT GENERATING FUNCTION (MGF)
𝓝(𝝁, 𝝈𝟐)
𝑡2
𝑀𝑈 𝑡 = 𝑒2
Recall,
1 (𝑥−𝜇) 2
𝑓𝑋 𝑥 = 𝑒− 2𝜎2
2𝜋𝜎2
Therefore,
𝑋 = 𝜇 + 𝜎𝑈
𝑀𝑋 𝑡 = 𝐸 𝑒𝑡𝑋
= 𝐸 𝑒𝑡 𝜇+𝜎𝑈

= 𝐸 𝑒 𝑡𝜇 𝑒 𝑡𝜎𝑈
= 𝑒 𝑡𝜇 𝐸[𝑒 𝑡𝜎𝑈 ]
= 𝑒𝑡𝜇𝑀𝑈[𝑡𝜎]
 MOMENT GENERATING FUNCTION (MGF)
𝓝(𝝁, 𝝈𝟐)
𝑀𝑋(𝑡) = 𝑒𝑡𝜇𝑀𝑈[𝑡𝜎]
(𝑡𝜎) 2
= 𝑒 𝑡𝜇 𝑒 2
𝑡𝜎 2
𝑀𝑋 𝑡 = exp + 𝜇𝑡
2
Differentiating,
𝑡𝜎 2
𝑀𝑋′ 𝑡 = (𝜇 + 𝑡𝜎2)exp + 𝜇𝑡
2
𝑡𝜎 2 𝑡𝜎 2
𝑀𝑋′′ 𝑡 = 𝜇 + 𝑡𝜎2 2exp + 𝜇𝑡 + 𝜎2exp + 𝜇𝑡
2 2
𝐸 𝑋 =𝜇
𝐸 𝑋 2 = 𝜇2 + 𝜎 2
𝑉𝑎𝑟 𝑋 = 𝜎2
TABLE OF MGF
Well-Known Distributions
TO UNDERSTAND A RANDOM VARIABLE

Just like its CDF or PMF (PDF), a random variable

is completely characterized by its MGF!

Department of Electronics and Communication Engineering

MOMENTS-RELATED FUNCTIONS

Characteristic Functions

Department of Electronics and Communication Engineering

CHARACTERISTIC FUNCTIONS

Generate “All” Moments In One-Shot

Mathematical Existence

Department of Electronics and Communication Engineering

CHARACTERISTIC FUNCTION (CF)
Definition
CHARACTERISTIC FUNCTION (CF)
Definition

• Note that because 𝑒 𝑗𝜔𝑥 = cos 𝜔𝑥 + 𝑗 sin 𝜔𝑥 ,

• 𝜙𝑋(𝜔) is generally a complex function; that is,
𝜙𝑋 𝜔 = 𝑈𝑋 𝜔 + 𝑗𝑉𝑋 (𝜔),
where
CHARACTERISTIC FUNCTION (CF)
Relationship with Fourier Transform
CHARACTERISTIC FUNCTION (CF)
Derivatives and Moments
CHARACTERISTIC FUNCTION (CF)
𝑩𝒆𝒓(𝒑)
CHARACTERISTIC FUNCTION (CF)
𝑩𝒊𝒏(𝒏, 𝒑)
 CHARACTERISTIC FUNCTION (CF)
𝑩𝒊𝒏(𝒏, 𝒑)
𝑛
𝜙𝑋 (𝜔) = 𝑝𝑒 𝑗𝜔 +1 −𝑝

Differentiating,
𝑛−1
𝜙𝑋′ 𝜔 = 𝑗𝑛 𝑝𝑒 𝑗𝜔 + 1 − 𝑝 𝑝𝑒 𝑗𝜔

𝑗𝐸 𝑥 = 𝜙𝑋′ 0 = 𝑗𝑛𝑝

𝑛−2 𝑛−1
2
𝜙𝑋′′ 𝜔 = 𝑛 𝑛 − 1 𝑝𝑒 𝑗𝜔 + 1 − 𝑝 𝑗𝑝𝑒 𝑗𝜔 − 𝑛 𝑝𝑒 𝑗𝜔 + 1 − 𝑝 𝑝𝑒 𝑗𝜔

𝜙𝑋′′ 0 = 𝑗 2 𝐸 𝑋2 = −(𝑛 𝑛 − 1 𝑝2 + 𝑛𝑝)

𝑉𝑎𝑟 𝑋 = 𝐸 𝑋2 − 𝐸 𝑋 2 = 𝑛𝑝(1 − 𝑝)
CHARACTERISTIC FUNCTION (CF)
𝑷𝒐𝒊(𝝀)
CHARACTERISTIC FUNCTION (CF)
𝑷𝒐𝒊(𝝀)

𝜙𝑋 𝜔 = exp 𝜆 𝑒 𝑗𝜔 − 1

Differentiating,
𝜙𝑋′ 𝜔 = 𝑗𝜆𝑒 𝑗𝜔 exp 𝜆 𝑒 𝑗𝜔 − 1
2
𝜙𝑋′′ 𝜔 = − 𝜆𝑒 𝑗𝜔 exp 𝜆 𝑒 𝑗𝜔 − 1 − 𝜆𝑒 𝑗𝜔 exp 𝜆 𝑒 𝑗𝜔 − 1

We have,
𝑗𝐸 𝑋 = 𝜙𝑋′ 0 = 𝑗𝜆
𝑗 2 𝐸 𝑋2 = 𝜙𝑋′′ 0 = −(𝜆2 + 𝜆)
𝑉𝑎𝑟 𝑋 = 𝜆2 + 𝜆 − 𝜆2 = 𝜆
CHARACTERISTIC FUNCTION (CF)
𝓝(𝟎, 𝟏)
CHARACTERISTIC FUNCTION (CF)
𝓝(𝟎, 𝟏)
 CHARACTERISTIC FUNCTION (CF)
𝓝(𝝁, 𝝈𝟐)
(𝑗𝜔) 2
𝜙𝑈 𝜔 = 𝑒 2
Recall,
1 (𝑥−𝜇) 2
𝑓𝑋 𝑥 = 𝑒− 2𝜎2
2𝜋𝜎2
Therefore,
𝑋 = 𝜇 + 𝜎𝑈
𝜙𝑋 𝜔 = 𝐸 𝑒 𝑗𝜔𝑋

= 𝐸 𝑒 𝑗𝜔 𝜇+𝜎𝑈

= 𝑒𝑗𝜔𝜇𝜙 𝑈 [𝑗𝜔𝜎]
 CHARACTERISTIC FUNCTION (CF)
𝓝(𝝁, 𝝈𝟐)
𝜙𝑋(𝜔) = 𝑒𝑗𝜔𝜇𝜙 𝑈 [𝑗𝜔𝜎]
𝑗𝜔𝜎 2
𝜙𝑋 𝜔 = exp + 𝑗𝜇𝜔
2
Differentiating,
𝜔𝜎 2
𝜙𝑋′ 𝜔 = (j𝜇 − 𝜔𝜎2)exp − + 𝑗𝜇𝜔
2
𝜔𝜎 2 𝜔𝜎 2
𝜙𝑋′′ 2 2
𝜔 = 𝑗𝜇 − 𝜔𝜎 exp − 2
+ 𝑗𝜇𝜔 − 𝜎 exp − + 𝑗𝜇𝜔
2 2
𝑗𝐸 𝑋 = 𝑗𝜇
𝑗 2 𝐸 𝑋2 = −(𝜇2 + 𝜎2)
𝑉𝑎𝑟 𝑋 = 𝜎2
RELATIONSHIP BETWEEN MGF AND CF
Direct relation

• Both MGF and CF generate raw (non-central) moments

• It is easy to verify that

𝜙𝑋 𝜔 = 𝜙𝑋(−𝑗𝜔)
TABLE OF MGF AND CF
Well-Known Distributions

Distribution MGF CF
Bernoulli 𝑋~𝐵𝑒𝑟(𝑝) 1 − 𝑝 + 𝑝𝑒𝜔 1 − 𝑝 + 𝑝𝑒 𝑗𝜔
Binomial 𝑋~𝐵𝑖𝑛(𝑛, 𝑝) 1 − 𝑝 + 𝑝𝑒 𝜔 𝑛 𝑛
1 − 𝑝 + 𝑝𝑒 𝑗𝜔
Poisson 𝑋~𝑃𝑜𝑖(𝜆) 𝑒𝜆 𝑒 𝜔 −1 𝑒 𝑗𝜔 −1
𝑒𝜆
𝜇 𝜔 +1 𝜎 2 𝜔 2 𝑗 𝜇 𝜔 −1 𝜎 2 𝜔 2
Gaussian 𝑋~𝒩 𝜇, 𝜎2 𝑒 2
𝑒 2
TO UNDERSTAND A RANDOM VARIABLE

Just like its CDF or PMF (PDF), a random variable

is completely characterized by its MGF or CF!
TRANSFORMATION ON A RANDOM VARIABLE

A Function of a RV is another RV
Derivative Rule for PDF
 TRANSFORMATION ON A RANDOM VARIABLE
Study of Systems

• Let 𝑋 be is a input random variable with known PDF and the output of the
system 𝑌 whose PDF needs to be calculated.
• This problem can be vied as "𝑏𝑙𝑎𝑐𝑘 𝑏𝑜𝑥" with input 𝑋, output 𝑌 and the
"𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝑐ℎ𝑎𝑟𝑒𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐" 𝑌 = 𝑔 𝑋 .

𝑋 𝑌
𝑌 = 𝑔(𝑋)
𝑓𝑌(𝑦)
𝑓𝑋 (𝑥)

• In general, 𝑋 can be continuous, mixed or discrete random variable.

• The transformation 𝑔 can be linear, nonlinear, segmented, staircase etc.
TRANSFORMATION ON A RANDOM VARIABLE
Recall A Property

• Lemma 1: If 𝑋 is a random variable and 𝑔 . is a function, then 𝑔(𝑋) is

also a random variable

• Examples: If 𝑋 is a RV, so is 𝑋 2 , 𝑋Τ4 , tan 𝑋 , cos−1 𝑋4 , …

• Since 𝑔(𝑋) is a RV – CDF, PDF, MGF and CF

• Question: Given that a RV 𝑋 has a CDF 𝐹𝑋(𝑥), how can one calculate the
CDF of the random variable 𝑌 = 𝑔(𝑋)? Denote it by 𝐹𝑌(𝑦)

• Question: Given that a RV 𝑋 has a PDF 𝑓𝑋 (𝑥), how can one calculate the
PDF of the random variable 𝑌 = 𝑔(𝑋)? Denote it by 𝑓𝑌 𝑦
TRANSFORMATION ON A RANDOM VARIABLE

Finding CDF from another CDF

M𝐨𝐧𝐨𝐭𝐨𝐧𝐢𝐜 𝐓𝐫𝐚𝐧𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧 𝒀 = 𝒈(𝑿)
3 Cases

• We consider following three cases:

• 𝑔 is monotonically 𝑖𝑛𝑐𝑟𝑒𝑎𝑠𝑖𝑛𝑔
• 𝑔 is monotonically 𝑑𝑒𝑐𝑟𝑒𝑎𝑠𝑖𝑛𝑔
• 𝑔 is non-monotonic
 CASE 1: 𝒈 is monotonically increasing
• A transform 𝑔 is monotonically increasing if 𝑔 𝑥1 < 𝑔 𝑥2 for any 𝑥1 < 𝑥2.
• We have 𝑋 = 𝑔−1(𝑌), exists and well behaved.
• In order to obtain the PDF of 𝑌, we first calculate the CDF.
𝐹𝑌 𝑦 = 𝑃 𝑌 ≤ 𝑦
𝐹𝑌 𝑦 = 𝑃 𝑔 𝑋 ≤ 𝑦
= 𝑃 𝑋 ≤ 𝑔−1(𝑦)
= 𝐹𝑋 𝑔−1(𝑦)
• Note that this can also be written as
𝐹𝑋(𝑥) = 𝐹𝑌 𝑔 (𝑥)
 CASE 1: 𝒈 is monotonically increasing

𝐹𝑌(𝑦) = 𝐹𝑋 𝑔−1(𝑦)
• Differentiating both the sides with respect to 𝑦 produces,
𝑑𝑔−1 𝑦
𝑓𝑌 𝑦 = 𝑓𝑋 𝑥
𝑑𝑦

or
𝐹𝑋(𝑥) = 𝐹𝑌 𝑔 (𝑥)
• Differentiating both the sides with respect to 𝑥 produces,
𝑑𝑦 𝑓𝑋 𝑥
𝑓𝑋 𝑦 = 𝑓𝑌 𝑥 ⇒ 𝑓𝑌 𝑦 =
𝑑𝑥 𝑑𝑦
𝑑𝑥
 CASE 2: 𝒈 is monotonically decreasing
• A transform 𝑔 is monotonically decreasing if 𝑔 𝑥1 > 𝑔 𝑥2 for any 𝑥1 < 𝑥2.
• Note that the event 𝑌 ≤ 𝑦 is equivalent to 𝑋 ≥ 𝑔−1(𝑦)
• In order to obtain the PDF of 𝑌, we first calculate the CDF.

𝐹𝑌 𝑦 = 𝑃 𝑌 ≤ 𝑦 = P X ≥ 𝑔−1 𝑦
= 1 − 𝐹𝑋(𝑔−1 𝑦 )
Monotonic Transformation
 Case 3: Nonmonotonic Transformation

• Consider a general function that is not necessarily monotonic.

• In this case, we cannot associate the event {𝑌 ≤ 𝑦} with events of the form
{𝑋 ≤ 𝑔−1(𝑦)} or {𝑋 ≥ 𝑔−1(𝑦)}
because the transformation is not monotonic.
 Case 3: Nonmonotonic Transformation

• We consider the event,

OBTAINING PDF OF 𝒀 = 𝒈(𝑿) FROM PDF OF 𝑿
Works With Most Functions Encountered

• Question: Given that a RV 𝑋 has a PDF 𝑓𝑋 (𝑥), how can one calculate the
PDF of the random variable 𝑌 = 𝑔(𝑋), directly? Denote it by 𝑓𝑌 (𝑦)
𝑓𝑋 (𝑥)
𝑓𝑌 𝑦 =
𝑑𝑦
𝑑𝑥
• This technique is called as a Jacobian transformation
• Note that there is absolute value of the derivative in the denominator
• Given that we know PDF 𝑓𝑌 (𝑦), we can directly calculate the CDF 𝐹𝑌(𝑦)
• From 𝑓𝑌 (𝑦), we can integrate to calculate 𝐹𝑌 𝑦 , if required
OBTAINING CDF OF 𝒀 = 𝒈(𝑿) FROM CDF OF 𝑿
Example

• Given that a RV 𝑋 has a CDF 𝐹𝑋(𝑥), calculate the CDF of the random
variable 𝑌 = 𝑋2.

𝐹𝑌 𝑦 = 𝑃 𝑌 ≤ 𝑦
= 𝑃 𝑋2 ≤ 𝑦
=𝑃 𝑋≤± 𝑦
=𝑃 𝑋 ≤ 𝑦
=𝑃 − 𝑦≤𝑋≤ 𝑦
𝐹𝑌 𝑦 = 𝐹𝑋 𝑦 − 𝐹𝑋(− 𝑦)

• The last statement is true because

𝑃 𝑎 ≤ 𝑋 ≤ 𝑏 = 𝐹𝑋 𝑏 − 𝐹𝑋(𝑎)
TRANSFORMATION ON A RANDOM VARIABLE

Finding PDF from another PDF

Rajini M.
Department of Electronics and Communication Engineering
OBTAINING PDF OF 𝒀 = 𝒈(𝑿) FROM PDF OF 𝑿
Example

• Given that a RV 𝑋 has a PDF 𝑓𝑋 (𝑥), calculate the PDF of the random
variable 𝑌 = 𝑋2.

𝑓𝑋 𝑦 𝑓𝑋 − 𝑦
𝑓𝑌 𝑦 = +
2| + 𝑦| 2| − 𝑦|

𝑓𝑋 𝑦 + 𝑓𝑋 − 𝑦
=
2 𝑦
 Transformation of a Discrete Random Variable

• Let 𝑋 is a discrete random variable and a transform 𝑔 is monotonic.

• There is one-to-one correspondence between 𝑋 and 𝑌 so that a set 𝑦𝑛
corresponds to the set 𝑥𝑛 through the equation
𝑦𝑛 = 𝑔(𝑥𝑛)
• Thus,
Transformation of a Discrete Random Variable
OBTAINING PMF OF 𝒀 = 𝒈(𝑿) FROM PMF OF 𝑿 DISCRETE
Example

• Given that a RV 𝑋 with values 𝑥 = −1, 0, 1, 2 and probabilities

0.1, 0.3, 0.4, 0.2respectively. Calculate the PDF of the random variable
𝑌 = 2 − 𝑋2 + 𝑋3/3.

𝑓𝑋 𝑥 = 0.1 𝛿 𝑥 + 1 + 0.3 𝛿 𝑥 + 0.4 𝛿 𝑥 − 1 + 0.2 𝛿 𝑥 − 2

𝑓𝑌 𝑦 = 0.3 𝛿 𝑦 − (2/3) + 0.4 𝛿 𝑦 − (4/3) + 0.3 𝛿 𝑦 − 2

TRANSFORMATION ON A RANDOM VARIABLE

Examples
Monotonic Transformation
OBTAINING PDF OF 𝒀 = 𝒈(𝑿) FROM PDF OF 𝑿
Works With Most Functions Encountered

• Question: Given that a RV 𝑋 has a PDF 𝑓𝑋 (𝑥), how can one calculate the
PDF of the random variable 𝑌 = 𝑔(𝑋), directly? Denote it by 𝑓𝑌 (𝑦)
𝑓𝑋 (𝑥)
𝑓𝑌 𝑦 =
𝑑𝑦
𝑑𝑥
• This technique is called as a Jacobian transformation
• Note that there is absolute value of the derivative in the denominator
• Given that we know PDF 𝑓𝑌 (𝑦), we can directly calculate the CDF 𝐹𝑌(𝑦)
• From 𝑓𝑌 (𝑦), we can integrate to calculate 𝐹𝑌 𝑦 , if required
 Example 3
Suppose 𝑋 is a Gaussian random variable with mean, 𝜇, and variance, 𝜎 2 . A new
random variable is formed according to 𝑌 = 𝑎𝑋 + 𝑏, where 𝑎 > 0. Find the PDF
of random variable 𝑌.
𝑌 − 𝑏 𝑑𝑦
𝑋= , =𝑎
𝑎 𝑑𝑥
1 𝑦−𝑏
𝑓𝑌 𝑦 = 𝑓𝑋
𝑎 𝑎
2
𝑦−𝑏
1 −𝜇
𝑎
𝑓𝑌 𝑦 = exp −
𝑎 2𝜋𝜎2 2𝜎2
1 𝑦 − (𝑏 + 𝑎𝜇) 2
= exp −
2𝜋(𝑎𝜎)2 2(𝑎𝜎)2
Note that the PDF of Y still has a Gaussian form. In this example, the transformation
did not change the form of the PDF; it merely changed the mean and variance.
 Example 4
Suppose 𝑋 is a exponential random variable with 𝑓𝑋 𝑥 = 2𝑒−2𝑥𝑢(𝑥). A new
random variable is formed according to 𝑌 = 𝑋3 . Find the PDF of random variable
𝑌.
𝑑𝑦
3
𝑋= 𝑌, = 3𝑥2
𝑑𝑥

2 −2 1
𝑓𝑌 𝑦 = 𝑦 3 exp −2𝑦3 𝑢(𝑦)
3
 Example 5
Suppose a phase angle Θ is uniformly distributed over (−𝜋/2, 𝜋/2), and the
transformation is 𝑌 = sin(Θ).
Solution:
• In general, 𝑦 = sin(𝜃) is not a monotonic transformation, but under the
restriction −𝜋/2 < 𝜃 < 𝜋/2, this transformation is indeed monotonically
increasing.
• With this transformation the resulting random variable, 𝑌, must take on values
in the range (−1, 1).
• Therefore, whatever PDF is obtained for Y, it must be understood that the PDF is
zero outside (−1, 1).
Example 5

𝑑𝑦
Θ = sin−1 𝑌 , = cos 𝜃
𝑑𝜃

1
𝑓𝑌 𝑦 =
𝜋cos(sin−1 𝑦)

1
𝑓𝑌 𝑦 = , −1 < 𝑦 < 1
𝜋 1 − 𝑦2
 Example 7
Suppose 𝑋 is a uniform random variable 𝑈(0,1). A new random variable is formed
according to 𝑌 = 𝑋 𝑛 . Find the PDF of random variable 𝑌.
𝑛
𝑋= 𝑌,
𝐹𝑌 𝑦 = 𝑃 𝑌 ≤ 𝑦 = 𝑃 𝑋 𝑛 ≤ 𝑦
𝑛
=𝑃 𝑋≤ 𝑦
𝑛
= 𝐹𝑋 𝑦
0, 𝑥 < 𝑎
𝑥−𝑎
𝐹𝑋 𝑥 = ,𝑎 ≤ 𝑥 ≤ 𝑏
𝑏−𝑎
1, 𝑥 > 𝑏
Example 7
0, 𝑥 < 0
𝑥 ,0 ≤ 𝑥 ≤ 1
𝐹𝑋 𝑥 =
1−0
1, 𝑥 > 1

𝑑
𝑓𝑌 𝑦 = 𝐹𝑌(𝑦)
𝑑𝑦

1 1−1
𝑓𝑌 𝑦 = 𝑦 𝑛 (𝑢 𝑛 − 1 − 𝑢(𝑛))
𝑛
 Example 8
If 𝑋 is continuous random variable with PDF 𝑓𝑋 𝑥 and suppose 𝑔 𝑥 is strictly
monotonic and differentiable then 𝑌 = 𝑔(𝑥) has a PDF
𝑑𝑔−1 𝑦
𝑓𝑌 𝑦 = 𝑓𝑋 𝑔−1 𝑦
𝑑𝑦
Find the PDF if 𝑌 = 𝑋 𝑛 in particular
Solution:
𝑛 𝑑𝑔−1 𝑦 1 1−1
𝑌= 𝑔 𝑥 = 𝑋𝑛 ⇒𝑋= 𝑌, = 𝑦 𝑛
𝑑𝑦 𝑛
1 1−1
𝑛
𝑓𝑌 𝑦 = 𝑦 𝑛 𝑓𝑋 𝑦
𝑛
If 𝑛 = 2, then
1
𝑓𝑌 𝑦 = 𝑓 𝑦
2 𝑦 𝑋
 Example 9
If 𝑋 is a normal random variable with mean 𝜇 and variance 𝜎 2 , then the random
variable 𝑌 = 𝑒 𝑋 is said to be 𝑙𝑜𝑔𝑛𝑜𝑟𝑚𝑎𝑙 random variable with parameters 𝜇 and
𝜎2.
Solution:
𝑑𝑔−1 𝑦 1
𝑌 = 𝑔 𝑥 = 𝑒 𝑋 ⇒ 𝑋 = log 𝑌 , =
𝑑𝑦 𝑦

1 log 𝑦 − 𝜇 2
𝑓𝑌 𝑦 = exp − ,y > 0
2𝜋𝜎𝑦 2𝜎2
CONDITIONAL DISTRIBUTIONS

Conditional CDF
Conditional PMF / PDF
CONDITIONAL DISTRIBUTIONS
Motivation and Definition

• Recall that for two events A and B, the conditional probability is defined as

𝑃 𝐴∩𝐵
𝑃 𝐴𝐵 = , 𝑃(𝐵) ≠ 0
𝑃(𝐵)
• The above notion can be extended to random variables too
• Let 𝐴 in the above equation be identified as the event 𝑋 ≤ 𝑥 for the
random variable 𝑋.
• The resulting probability 𝑃 𝑋 ≤ 𝑥 𝐵 is defined as the conditional
distribution function of 𝑋, denoted as 𝐹𝑋|𝐵(𝑥)

𝑃 𝑋 ≤𝑥∩𝐵
𝐹𝑋|𝐵(𝑥) = 𝑃 𝑋 ≤ 𝑥 𝐵 =
𝑃(𝐵)
CONDITIONAL DISTRIBUTIONS
Definition

• The conditional distribution function of 𝑋,

𝑃 𝑋 ≤𝑥∩𝐵
𝐹𝑋|𝐵(𝑥) = 𝐹𝑋 𝑥 𝐵 = 𝑃 𝑋 ≤ 𝑥 𝐵 =
𝑃(𝐵)
• 𝑃 𝑋 ≤ 𝑥 ∩ 𝐵 is the joint event 𝑋 ≤ 𝑥 ∩ 𝐵.
• This consists of all the outcomes 𝒔 such that
𝑋 𝑠 ≤ 𝑥 and 𝒔 ∈ 𝐵
CONDITIONAL DISTRIBUTION FUNCTION
Properties

1. Conditional CDF is monotonically non-decreasing function in 𝑥, i.e.,

𝐹𝑋|𝐵 𝑥2 ≥ 𝐹𝑋|𝐵 𝑥1 , if 𝑥2 > 𝑥1, ∀ 𝑥1, 𝑥2 ∈ ℝ

2. lim 𝐹𝑋|𝐵 𝑥 = 0 𝑎𝑛𝑑 lim 𝐹𝑋|𝐵 𝑥 = 1.

𝑥→−∞ 𝑥→∞
3. 𝐹𝑋|𝐵 𝑥 is right continuous in 𝑥, for instance

1
lim 𝐹𝑋|𝐵 𝑥+ = 𝐹𝑋|𝐵 𝑥 .
𝑁→∞ 𝑁

4. 𝑃 𝑥1 < 𝑋 ≤ 𝑥2|𝐵 = 𝐹𝑋|𝐵 𝑥2 − 𝐹𝑋|𝐵(𝑥1)

CONDITIONAL PMF
Definition

• The conditional PMF can be naturally defined as

𝑃 𝑋 = 𝑥1, 𝑋 = 𝑥2 𝑝𝑋(𝑥1, 𝑥2)

𝑝𝑋 𝑥1 𝑥2 = 𝑃 𝑋 = 𝑥1 𝑋 = 𝑥2 = =
𝑃 𝑋 = 𝑥2 𝑝𝑋(𝑥2)

• Such that 𝑝𝑋(𝑥2) = 𝑃 𝑋 = 𝑥2 > 0

• 𝑝𝑋 𝑥1 𝑥2 is a valid PMF
CONDITIONAL PDF
Motivated From Conditional CDF

• It is often more convenient to work with conditional PDF rather than a

conditional CDF.
• The conditional PDF of a random variable 𝑋 conditioned on the event 𝐵 is
𝑃 (𝑥 ≤ 𝑋 < 𝑥 + 𝜖|𝐵)
𝑓𝑋|𝐵 (𝑥) = lim
𝜖 →0 𝜖
CONDITIONAL PROBABILITY DENSITY FUNCTION (PDF)
Properties
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

Two boxes have red, green, and blue balls in them; the number of balls of each
color is given in Table. Our experiment will be to select a box and then a ball
from the selected box. One box (number 2) is slightly larger than the other,
causing it to be selected more frequently.
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

• Let 𝐵2 be the event "select the larger box" while 𝐵1 is the event "select the

smaller box."

• Assume 𝑃(𝐵1) = 2/10 and 𝑃(𝐵2) = 8/10.

• 𝐵1and 𝐵2 are mutually exclusive and 𝐵1 ∩ 𝐵2 is the certain event, since some

box must be selected

• Therefore, 𝑃 𝐵1 + 𝑃 𝐵2 = 1.
 CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

• Now define a discrete random variable 𝑋 to have values

𝑥1 = 1, 𝑥2 = 2, 𝑎𝑛𝑑 𝑥3 = 3 when a red, green, or
blue ball is selected.
• Let 𝐵 be an event equal to either 𝐵1 𝑜𝑟 𝐵2.

5 80
𝑃 𝑋 = 1 𝐵 = 𝐵1 = 𝑃 𝑋 = 1 𝐵 = 𝐵2 =
100 150
35 60
𝑃 𝑋 = 2 𝐵 = 𝐵1 = 𝑃 𝑋 = 2 𝐵 = 𝐵2 =
100 150
60 10
𝑃 𝑋 = 3 𝐵 = 𝐵1 = 𝑃 𝑋 = 3 𝐵 = 𝐵2 =
100 150
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

• The conditional density function 𝑓𝑋 𝑥 𝐵1 and 𝑓𝑋 𝑥 𝐵2

5 35 60
𝑓𝑋|𝐵1 (𝑥) = 𝛿 𝑥−1 + 𝛿 𝑥−2 + 𝛿 𝑥−3
100 100 100
80 60 10
𝑓𝑋|𝐵2 (𝑥) = 𝛿 𝑥−1 + 𝛿 𝑥−2 + 𝛿(𝑥 − 3)
150 150 150
• The conditional distribution function 𝐹𝑋 𝑥 𝐵1 and 𝐹𝑋 𝑥 𝐵2
5 35 60
𝐹𝑋|𝐵1 (𝑥) = 𝑢 𝑥−1 + 𝑢 𝑥−2 + 𝑢 𝑥−3
100 100 100
80 60 10
𝐹𝑋|𝐵2 (𝑥) = 𝑢 𝑥−1 + 𝑢 𝑥−2 + 𝑢(𝑥 − 3)
150 150 150
 DISTRIBUTION ANF DENSITY FUNCTION
Examples

• For comparison, lets find the PDF and CDF by determining the probabilities
𝑃(𝑋 = 1), 𝑃(𝑋 = 2), 𝑎𝑛𝑑 𝑃(𝑋 = 3).
• These are found from the total probability theorem
𝑃 𝑋 = 1 = 𝑃 𝑋 = 1 𝐵1 𝑃 𝐵1 + 𝑃 𝑋 = 1 𝐵2 𝑃(𝐵2)
5 2 80 8
= ∗ + ∗ = 0.43
100 10 150 10
35 2 60 8
𝑃 𝑋 =2 = ∗ + ∗ = 0.390
100 10 150 10
60 2 10 8
𝑃 𝑋 =3 = ∗ + ∗ = 0.173
100 10 150 10
DISTRIBUTION ANF DENSITY FUNCTION
Examples
• The PDF and CDF are respectively
𝑓𝑋 (𝑥) = 0.437𝛿 𝑥 − 1 + 0.390𝛿 𝑥 − 2 + 0.173𝛿(𝑥 − 3)
𝐹𝑋(𝑥) = 0.437𝑢 𝑥 − 1 + 0.390𝑢 𝑥 − 2 + 0.173𝑢 𝑥 − 3
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

Suppose a RV is uniformly distributed over the interval [0, 1) so that its CDF
is given by

1
Find the conditional CDF of 𝑋 given that 𝑋 < .
2
Solution:
• The event 𝐴 = {𝑋 ≤ 1/2} is a numerical condition on the RV
• Using the definition of conditional CDF
𝑃 𝑋 ≤ 𝑥, 𝑋 ≤ 1/2
𝐹𝑋|{𝑋<1/2} 𝑥 =
𝑃(𝑋 ≤ 1/2)
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples

Case I: 𝑥 < 0,
• The intersection 𝑋 ≤ 𝑥 𝑎𝑛𝑑 {𝑋 ≤ 1/2} is simply the event 𝑋 ≤ 𝑥 ,
• But 𝑃 𝑋 ≤ 𝑥 = 0
𝑃 𝑋 ≤ 𝑥, 𝑋 ≤ 1/2 𝑃 𝑋≤𝑥
𝐹𝑋|{𝑋<1/2} 𝑥 = = =0
𝑃(𝑋 ≤ 1/2) 𝑃(𝑋 ≤ 1/2)
Case II: 0 ≤ 𝑥 ≤ 1/2
𝑃 𝑋 ≤ 𝑥 ∩ 𝑋 ≤ 1/2 𝑃 𝑋≤𝑥 𝑥
𝐹𝑋|{𝑋<1/2} 𝑥 = = = = 2𝑥
𝑃(𝑋 ≤ 1/2) 𝑃(𝑋 ≤ 1/2) 1/2
Case III: 𝑥 > 1/2
𝑃 𝑋 ≤ 𝑥 ∩ 𝑋 ≤ 1/2 𝑃 𝑋 ≤ 1/2
𝐹𝑋|{𝑋<1/2} 𝑥 = = =1
𝑃(𝑋 ≤ 1/2) 𝑃(𝑋 ≤ 1/2)
CONDITIONAL DISTRIBUTION ANF DENSITY FUNCTION
Examples
THANK YOU