Unit 3

Random Variable and Random Process

Random Variables (RV)

• It is numerical description of the outcome of a random experiment.

• A random variable is a numerically valued function defined over a

sample space.

There are two types of random variables:

1. Discrete Random Variable (DRV): If states only countable number
of discrete values then it is called DRV.

2. Continuous Random Variable (CRV): If can assume any value then

it is called CRV.
Cumulative Distribution Function (CDF)
• CDF is the probability of all possible values in the total available range.

• It contains all probabilistic information of random variable and is given by

𝐹 𝑋 ( 𝑥 ) =Pr [ 𝑋 ≤ 𝑥 ] ; − ∞ ≤ 𝑥 ≤ ∞
-- representing to probability
Example: 𝐹 𝑋 ( 𝑥)
1
0.75

0.50

0.25
𝑥
0 1 2 3 4
Properties of CDF:

• No possible event

• All possible events

This property indicates that CDF is a non-decreasing function of .

Probability Density(Distribution) Function (PDF)

A CRV can’t be described by any mathematical equation but we can definitely describe the
probability of CRV lying within a range around any possible value by a mathematical function.

This analytical function is known as PDF and is given by

𝑑 𝐹 𝑋 (𝑥 )
𝑝𝑋 ( 𝑥 )=
𝑑𝑥
Properties of PDF:

• Since CDF is always a non-decreasing function of , so its derivative will always be

greater than or equal to zero.

• Area under the PDF curve is always unity, i.e.,

∞

∫ 𝑝 𝑋 ( 𝑥 ) 𝑑𝑥=1
−∞
• Mean or First Moment:-
Mean is a measure of where distribution is centered and is given by

represents to the expectation.

Example: Find the mean of the following rectangular pulse:

𝑝𝑋(𝑥)
1
𝑏− 𝑎

𝑥
0 𝑎 𝑏
 • Mean Square Value:-
The mean square value of a random variable is given by expectation of .
∞
𝔼 [ 𝑋 ]= ∫ 𝑥 ⋅ 𝑝 𝑋 ( 𝑥 ) 𝑑𝑥
2 2

−∞
Also known as total average power

Example: Find the mean square value of the following rectangular pulse:

𝑝𝑋(𝑥)
1
𝑏− 𝑎

𝑥
0 𝑎 𝑏
• Variance:-
• It measures the spread of distribution about its mean.
• The less the spread, smaller is the variance and vice-versa.
𝑝 𝑋1 ( 𝑥) 𝑝 𝑋 2( 𝑥)

𝜎 2𝑥 1 𝜎 2𝑥 2

𝑚𝑥 1 𝑥 𝑚𝑥 2 𝑥

The variance of a random variable is given by

∞
𝜎 =𝔼 [ ( 𝑋 −𝑚 𝑥 ) ]= ∫ ( 𝑋 −𝑚 𝑥 ) ⋅𝑝 𝑋 ( 𝑥 ) 𝑑𝑥
2 2 2
𝑥
−∞
• Properties of expectation operation:

For Example:
Problems:

1. Consider a random variable which is uniformly distributed between to .

Calculate (i) mean, (ii) mean square value, and (iii) variance.

𝑝𝑋(𝑥)

1
2𝜋

0 𝑥
2𝜋

2. A random variable has CDF given by .

Find:
 ∞ ∞
𝑑
Pr [ 𝑋 ≥100 ] =∫ 𝑝 𝑋 ( 𝑥 ) 𝑑𝑥= ∫ ∞
𝐹 𝑋 ( 𝑥 ) 𝑑𝑥=𝐹 𝑋 ( 𝑥 ) |100 = 𝐹 𝑋 ( ∞ ) − 𝐹 𝑋 ( 100 )
100 100 𝑑𝑥

¿ 1− (1− 𝑒
− 0.01 ∗100
)= 1
𝑒

ADC - IIIT Bhopal 12

 Homework Problem
3. Consider a random variable with PDF given by

Calculate the relation between and ?

 • Random Process

The time domain representation of a random variable is known as random process.

There are two types of random process:

1. Stationary Random Process:

A random process is set to be stationary if its statistical properties are not the function of time.
Statistical properties, i.e., mean, variance, autocorrelation function (ACF), etc.

It is further of two types

1. Strictly stationary random process: If the random process has same statistical property, i.e.,
mean, variance, autocorrelation function (ACF), etc. at any instant of time then is set to be
stationary in strict sense.

2. Wide sense stationary (WSS) random process: If only mean and autocorrelation function
are stationary then is set to be WSS.
2. Non-Stationary Random Process:

A random process is set to be non-stationary if its statistical properties are the function of
time.
• Autocorrelation Function (ACF) :-

The autocorrelation function of a random process is given by

𝑅 𝑋 ( 𝜏 )=𝔼 [ 𝑋 (𝑡 ) ⋅ 𝑋 ( 𝑡 +𝜏 ) ]
𝑅 𝑋 1 (𝜏 ) 𝑅𝑋 2 (𝜏 )
Slowly varying
Rapidly fluctuating random process
random process

𝑥 𝑥
𝜏=0 𝜏=0
• Physical significance: The physical significance of random process is that it provides the
mean of describing the independence of random process at time second apart.
• It is therefore clear that more rapidly random process changes with time, more rapidly
will the ACF decreases from its maximum value.
Example:
Consider a random process is given by where is uniformly distributed
between and . Calculate .

𝑅 𝑋 ( 𝜏 )=𝔼 [ 𝑋 (𝑡 ) ⋅ 𝑋 ( 𝑡 +𝜏 ) ]

∵0≤𝜃≤2 𝜋
• Properties of autocorrelation function:

• Correlation has its maximum value at , i.e.,

• is always an even function of , i.e.,

• Mean square value of a random process can be obtained from by simply

putting .
• Ergodic Random Process

A random process is set to be ergodic

when time average of a random process becomes equal to ensemble mean.

• Properties of Ergodic Random Process:

• The ensemble mean gives the DC component of the random process.

• is the DC power of the ergodic random process.
• The mean square value of a random process is the total average power of
the ergodic random process.

At , 𝑇 𝑇
1 1
𝑅 𝑋 ( 0 ) = ∫ 𝑋 ( 𝑡 ) ⋅ 𝑋 ( 𝑡+𝜏 ) 𝑑𝑡 = ∫ 𝑋 (𝑡 ) 𝑑𝑡
2
2𝑇 − 𝑇 2𝑇 −𝑇
ADC - IIIT Bhopal 20
 After solving, we get

𝜎 𝑥 =𝔼 [ 𝑥 ] − 𝑚𝑥
2 2 2

DC Power
Variance or
AC Power
Total Average
Power

ADC - IIIT Bhopal 21

Problem:
A stationary random process has a autocorrelation function given by
− 4|𝜏|
( )
𝑅 𝑋 𝜏 =20+5 cos 2 𝜏+10 𝑒
Evaluate:

(a) Total average power = 20 + 5 + 10 = 35 W

(b) DC power 20 W
(c) AC power = 5 + 10 = 15 W
(d) Mean 𝑚2𝑥 =20 ⇒ 𝑚 𝑥 = √ 20
(e) Variance 15 W
(f) RMS value of random process
(Mean square value) 𝑋 𝑟𝑚𝑠= √ 35

ADC - IIIT Bhopal 22

Power Spectral Density (PSD):

---- Power per unit BW

PSD = Watt/rad/sec or Watt/Hz

()

𝑓
−𝑓𝑚 𝑓𝑚

ADC - IIIT Bhopal 23

• Properties of PSD:

• or

• or

ADC - IIIT Bhopal 24

Gaussian/Normal PDF

𝑝 𝑋 ( 𝑥 )=
1
√2 𝜋 𝜎 2
𝑥
⋅ exp
[ − ( 𝑥 −𝑚 𝑥 )2
2 𝜎 2𝑥 ]
𝑝𝑋( 𝑥)

𝑥
𝑚𝑥

ADC - IIIT Bhopal 25

Central Limit Theorem:

Central Limit Theorem states that probability density

of a sum of N independent random variables tends
to approach a Gaussian density when the number N
increases.

• This theorem is applicable even when the individual random variable are not Gaussian.

• For large number of N (number of samples), sum of their PDFs forms a normal (Gaussian) curve.

• Maximum value of the resultant PDF can be obtained at its mean.

ADC - IIIT Bhopal 26