0% found this document useful (0 votes)

6 views65 pages

QTA 3 - Common Univariate Random Variables

Uploaded by

Ravishakumar Shetty

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

6 views65 pages

QTA 3 - Common Univariate Random Variables

Uploaded by

Ravishakumar Shetty

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 65

1 of 65

QTA 3 -
COMMON UNIVARIATE
RANDOM VARIABLES
Source - Global Association of Risk Professionals (GARP)

Micky Midha
BE, CFA®, FRM, LLB www.midhafin.com
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
2 of 65

LEARNING OBJECTIVES www.midhafin.com

• Distinguishing the key properties among the following distributions :
 Uniform distribution,

 Bernoulli distribution,

 Poisson distribution,

 Normal distribution,

 Lognormal distribution,

 Chi-squared distribution,

 Student’s t distribution, and

 F distribution

• Identifying common occurrences of each distribution.

• Describe a mixture distribution and explain the creation and characteristics of

mixture distributions.
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
3 of 65

INTRODUCTION www.midhafin.com
• There are three common discrete distributions:

 The Bernoulli is a general purpose distribution that is typically used to model binary events.

 The binomial distribution describes the sum of 𝑛 independent Bernoulli random variables.

 The Poisson distribution is commonly used to model hazard rates, which count the number
of events that occur in a fixed unit of time (e.g., the number of corporations defaulting in
the next quarter).

____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
4 of 65

INTRODUCTION www.midhafin.com

• There is a wider variety of continuous distributions used by risk managers.

• The most basic is a uniform distribution, which serves as a foundation for all random variables.

• The most widely used distribution is the normal, which is used for tasks such as modeling
financial returns and implementing statistical tests.

• Many other frequently used distributions are closely related to the normal. These include
 Student’s 𝑡,
 Chi-square 𝜒 , and
 𝐹,

• Mixture distributions are built using two or more distinct component distributions. Mixtures
can be used to build distributions that match important features of financial data. For example,
mixing two normal random variables with different variances produces a random variable
that has a larger kurtosis than either of the mixture components.

BERNOULLI DISTRIBUTION www.midhafin.com

• The Bernoulli is a discrete distribution for random variables that produces one of two values: 0
or 1. It applies to any problem with a binary outcome (e.g., bull and bear markets, corporate
defaults, or the classification of fraudulent transactions).

• The Bernoulli distribution depends on a single parameter, 𝑝, which is the probability that a
success is observed.

• Suppose that 𝑌 is a Bernoulli random variable with parameter 𝑝. This can be expressed as
𝑌~𝐵𝑒𝑟𝑛𝑜𝑢𝑙𝑙𝑖 𝑝

• The mean of 𝑌 is
𝐸𝑌 𝑝 1 1 𝑝 0 𝑝,

• The Variance of 𝑌 is
𝑉𝑌 𝐸𝑌 𝐸𝑌 𝑝 1 1 𝑝 0 𝑝 𝑝 1 𝑝 𝑝𝑞
where
𝑞 1 𝑝 is the failure probability.
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
6 of 65

BERNOULLI DISTRIBUTION www.midhafin.com

• The probability mass function (PMF) of a • The CDF of a 𝐵𝑒𝑟𝑛𝑜𝑢𝑙𝑙𝑖 𝑝 is a step
𝐵𝑒𝑟𝑛𝑜𝑢𝑙𝑙𝑖 𝑝 is function with three values:
𝑓 𝑦 𝑝 1 𝑝 0 𝑦 0
𝐹 𝑦 1 𝑝 0 𝑦 1
This function only produces two values: 1 𝑦 1
𝑝, when 𝑦 1
1 𝑝, when 𝑦 0

BINOMIAL DISTRIBUTION www.midhafin.com

• A binomial random variable measures the total number of successes from 𝑛 independent
Bernoulli random variables, where each has a probability of success equal to 𝑝. Hence, binomial
distributions are used to model counts of independent events

A binomial distribution has two parameters:

1. 𝑛, the number of independent experiments; and
2. 𝑝, the probability that each experiment is successful.

• If 𝑛 variables 𝑋 ∼ 𝐵𝑒𝑟𝑛𝑜𝑢𝑙𝑙𝑖 𝑝 are independent, then a binomial with parameters 𝑛 and 𝑝 is

defined as

𝑌 𝑋

and expressed as
𝑌 𝐵 𝑛, 𝑝

BINOMIAL DISTRIBUTION www.midhafin.com

• The mean of a binomially distributed random variable is given by
𝐸𝑌 𝑛𝑝
The variance of 𝑌 is
𝑉𝑌 𝑛𝑝 1 𝑝

• The skewness of a binomial depends on 𝑝, with small values producing right-skewed

distributions.

• The normal distribution provides a convenient approximation to the binomial if both

𝑛𝑝 10
and
𝑛 1 𝑝 10

BINOMIAL DISTRIBUTION www.midhafin.com

• The PMF of a 𝐵 𝑛, 𝑝 is • The CDF is the sum of the cumulated PMF
𝑛 between 0 and 𝑦
𝑓 𝑦 𝑝 1 𝑝 ⌊ ⌋
𝑦 𝑛
where 𝐹 𝑦 𝑝 1 𝑝
𝑖
𝑛 𝑛!
𝐶
𝑦 𝑦! 𝑛 𝑦 !

BINOMIAL DISTRIBUTION www.midhafin.com

• The probability that each stock will go up during an expansion is 0.80. An investor has 9 such
stocks and the event of going up for any particular stock is independent of the other stocks
going up. If 𝑋 denote the number stocks that go up, what is the mean, variance and standard
deviation of 𝑋? What is the probability that zero, five, seven or all stocks will go up?

POISSON DISTRIBUTION www.midhafin.com

• Poisson random variables are used to measure counts of events over fixed time spans. For
example, one application of a Poisson is to model the number of loan defaults that occur each
month.

• Poisson random variables are always non-negative and integer-valued. The Poisson distribution
has a single parameter, which is called the hazard rate and expressed as 𝜆, that signifies the
average number of events per interval.

• The mean and variance of 𝑌 ∼ 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 𝜆 are simply:

𝐸𝑌 𝑉𝑌 𝜆

POISSON DISTRIBUTION www.midhafin.com

• If the rate at which events occur over time is constant, and the probability of any one event
occurring is independent of all other events, then we say that the events follow a Poisson
process, where:
𝜆𝑡
𝑃𝑌 𝑛 𝑒
𝑛!
where 𝑡 is the time elapsed.

• A useful feature of the Poisson (one that is uncommon among distributions) is that it is
infinitely divisible.
If 𝑋 ∼ 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 𝜆 , and 𝑋 ∼ 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 𝜆 are independent, and
𝑌 𝑋 𝑋 , then 𝑌 ∼ 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 𝜆 𝜆 .
This feature makes the Poisson well-suited to work with timeseries data, because summing the
number of events in a sampling interval (e.g., a week, month, or quarter) does not change the
distribution.

POISSON DISTRIBUTION- EXAMPLE www.midhafin.com

• If the defaults in a portfolio of bonds follow a Poisson distribution and the expected number of
defaults each month is five, then find the probability that there are exactly four defaults during
the next month? Also find the probability that there are exactly five defaults over the next two
months?
𝜆𝑡
𝑃𝑌 𝑛 𝑒
𝑛!

POISSON DISTRIBUTION www.midhafin.com

• The PMF of a Poisson random variable is • The CDF of a Poisson is defined as the sum
𝜆 e of the PMF for values less than the input
𝑓 𝑦 ⌊ ⌋
𝑦! 𝜆
𝐹 𝑦 e
𝑖!

UNIFORM DISTRIBUTION www.midhafin.com

• The simplest continuous random variable is a uniform random variable. A uniform distribution
assumes that any value within the range [𝑎, 𝑏] is equally likely to occur.

• The PDF of a uniform random variable does not depend on 𝑦, because all values are equally
likely.

• The CDF returns the cumulative probability of observing a value less than or equal to the
argument. The CDF is 0 to the left of 𝑎, which is the smallest value that could be produced,
linearly increases from 𝑎 𝑡𝑜 𝑏, and then is 1 above 𝑏.

• When 𝑎 0 and 𝑏 1, the distribution is called the standard uniform. Any uniform random
variable can be constructed from a standard uniform 𝑈 using the relationship:
𝑈 𝑎 𝑏 𝑎 𝑈,
where 𝑎 and 𝑏 are the bounds of 𝑈 .

UNIFORM DISTRIBUTION www.midhafin.com

• The PDF of a uniform distribution is • The CDF of a uniform distribution is
0 𝑖𝑓 𝑦 𝑎 0 𝑖𝑓 𝑦 𝑎

1 𝑦 𝑎
𝑓 𝑦 𝑖𝑓 𝑎 𝑦 𝑏 𝐹 𝑦 𝑖𝑓 𝑎 𝑦 𝑏
𝑏 𝑎 𝑏 𝑎
0 𝑖𝑓 𝑦 𝑏 1 𝑖𝑓 𝑦 𝑏

UNIFORM DISTRIBUTION www.midhafin.com

• The mean of a uniform random variable 𝑌 ∼ 𝑈 𝑎, 𝑏 is the mid-point of the support
𝑎 𝑏
𝐸𝑌
2

• The variance depends only on the width of the support:

𝑏 𝑎
𝑉𝑌
12

• The mean and variance of a standard uniform random variable are ⁄ and ⁄ , respectively.

• The probability that 𝑌 falls into an interval with lower bound 𝑙 and upper bound 𝑢 i.e.,
𝑚𝑖𝑛 𝑢, 𝑏 𝑚𝑎𝑥 𝑙, 𝑎
𝑃 𝑙 𝑌 𝑢
𝑏 𝑎
which simplifies to
𝑢 𝑙
𝑏 𝑎
when both 𝑙 𝑎 𝑎𝑛𝑑 𝑢 𝑏
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
23 of 65

UNIFORM DISTRIBUTION www.midhafin.com

𝑚𝑖𝑛 𝑢, 𝑏 𝑚𝑎𝑥 𝑙, 𝑎
𝑃 𝑙 𝑌 𝑢
𝑏 𝑎

UNIFORM DISTRIBUTION www.midhafin.com

• The current (in mA) measured in a piece of copper wire is known to follow a uniform
distribution over the interval [1,11]. Find PDF and CDF of the random variable 𝑋 representing
the current. Calculate the mean and variance of the distribution.

UNIFORM DISTRIBUTION www.midhafin.com

NORMAL DISTRIBUTION www.midhafin.com

• The normal distribution is the most commonly used distribution in risk management. It is also
commonly referred to as a Gaussian distribution (after Carl Friedrich Gauss) or a bell curve
(which reflects the shape of the PDF). The normal distribution is popular for many reasons.
 Many continuous random variables are approximately normally distributed.
 The distribution of many discrete random variables can be well approximated by a normal.
 The normal distribution plays a key role in the Central Limit Theorem (CLT), which is
widely used in hypothesis testing (i.e., the process where data is used to determine the
truthfulness of an objective statement).
 Normal random variables are infinitely divisible, which makes them suitable for simulating
asset prices in models that assume that prices are continuously evolving.
 The normal is closely related to many other important distributions, including the
Student’s 𝑡, the 𝜒 , and the 𝐹.
 The normal is closed (i.e., weighted sums of normal random variables are normally
distributed) under linear operations.
 Estimators derived under the assumption that the underlying observations are normally
distributed often have simple closed forms.

NORMAL DISTRIBUTION www.midhafin.com

• It is completely described by its mean, 𝜇, and variance,𝜎 , stated as 𝑌~𝑁 𝜇, 𝜎
𝐸𝑌 𝜇 and 𝑉 𝑌 𝜎

• Skewness=0, thus 𝑃 𝑋 𝜇 𝑃𝜇 𝑋 0.5

• Kurtosis=3. This kurtosis is often used as a benchmark when assessing whether another
distribution is heavy/fat-tailed.

• The normal can generate any value in ( ∞, ∞), although it is unlikely to observe values more
than 3𝜎 away from the mean. In fact, values more than 3𝜎 away from the mean are expected in
only one in 370 realizations of a normal random variable.

NORMAL DISTRIBUTION www.midhafin.com

• The PDF of a normal distribution is • The CDF of 𝑌 is
1 𝑦 𝜇 𝑦 𝜇
𝑓 𝑦 exp 𝐹 𝑦 Φ
2𝜋𝜎 2𝜎 𝜎

NORMAL DISTRIBUTION www.midhafin.com

• the sums of independent normally distributed random variables are also normally distributed. If
𝑋 ∼ 𝑁 𝜇 , 𝜎 and 𝑋 ∼ 𝑁 𝜇 , 𝜎 are independent, and
𝑌 𝑋 𝑋 , then
𝑌∼𝑁 𝜇 𝜇 ,𝜎 𝜎
This property simplifies describing log returns at different frequencies; if daily log returns are
independent and normal, then weekly and monthly returns are as well.

• Tails get thinner and thinner and extend up to infinity.

Mean=Median=Mode

NORMAL DISTRIBUTION – CONFIDENCE INTERVALS www.midhafin.com

• 68% confidence interval is
𝜇 𝜎 to 𝜇 𝜎

• 90% confidence interval is

𝜇 1.645𝜎 to 𝜇 1.645𝜎

• 95% confidence interval is

𝜇 1.96𝜎 to 𝜇 1.96𝜎

• 99% confidence interval is

𝜇 2.58𝜎 to 𝜇 2.58𝜎

NORMAL DISTRIBUTION – CONFIDENCE INTERVALS www.midhafin.com

EXAMPLE –

• The return on a stock is 12% per annum and the standard deviation of annual returns is 15%.
Assuming that the returns follow a normal distribution, find the 90% confidence interval for the
stock return over this year.

STANDARD NORMAL DISTRIBUTION www.midhafin.com

• A standard normal distribution (or a z-distribution) is a normal distribution which has been
standardized to have a mean of zero and a standard deviation of one. It is stated as 𝑋~𝑁 0,1

• It is also common to use 𝜙 𝑧 to denote the standard normal PDF and Φ 𝑧 to denote the
standard normal CDF.

PDF is
1
𝜙 e
2𝜋
• The z-value is the number of standard deviations a given observation is away from the mean.
Standardization is converting an observed value 𝑦 to its z value by using the formula:
𝑦 𝜇
z
𝜎

STANDARD NORMAL DISTRIBUTION - EXAMPLE www.midhafin.com

• Consider out last example where the annual return on a stock is normally distributed with a mean
of 12% per annum and a standard deviation of 15%. Find the z-values corresponding to the
returns -3% and 19.5%

STANDARD NORMAL DISTRIBUTION - EXAMPLE www.midhafin.com

STANDARD NORMAL DISTRIBUTION AND PROBABILITIES www.midhafin.com

• A standard normal table, or z-table, is a mathematical table for the values of Φ, which are values
of the cumulative distribution function of the normal distribution, denoted by Φ 𝑍 𝑃𝑍 𝑧
By symmetry of the normal distribution Φ 𝑍 1 Φ 𝑍

• The values in the z-table are the probabilities of observing a z-value that is less than a given value
(or 𝑃 𝑍 𝑧 ). It can also be used to find the probability that a z-value is observed above (by
using 𝑃 𝑍 𝑧 1 𝑃 𝑍 𝑧 ), or between values on the standard normal distribution, and by
extension, any normal distribution.

STANDARD NORMAL DISTRIBUTION AND PROBABILITIES www.midhafin.com

STANDARD NORMAL TABLE – EXAMPLE 1 www.midhafin.com

• Consider out last example where the annual return on a stock is normally distributed with a mean
of 12% per annum and a standard deviation of 15%. Find the probability that the return is less
than or equal to 18%.

STANDARD NORMAL TABLE – EXAMPLE 2 www.midhafin.com

STANDARD NORMAL TABLE – EXAMPLE 3 www.midhafin.com

APPROXIMATING DISCRETE RANDOM VARIABLES www.midhafin.com

• A normal distribution can approximate a binomial random variable if both
𝑛𝑝 10 and 𝑛 1 𝑝 10
When these conditions are satisfied, a binomial has either many independent experiments or a
probability that is not extreme (or both), and so the PMF is nearly symmetric and well
approximated by a
𝑁 𝑛𝑝, 𝑛𝑝 1 𝑝

• The Poisson can also be approximated by a normal random variable. When 𝜆 is large, then a
Poisson 𝜆 can be well approximated by a Normal 𝜆, 𝜆 . This approximation is commonly
applied when 𝜆 1000.

LOGNORMAL DISTRIBUTION www.midhafin.com

• A variable 𝑌 is said to be log-normally distributed if the natural logarithm of 𝑌 is normally
distributed. In other words, if 𝑋 𝑙𝑛 𝑌, then is log-normally distributed if and only if 𝑋 is
normally distributed. Alternatively, a log-normal can be defined

𝑌 𝑒
where 𝑋 ∼ 𝑁 𝜇, 𝜎 .

• An important property of the log-normal distribution is that 𝑌 can never be negative, whereas
𝑋 can be negative because it is normally distributed. This log-normal property can be desirable
when constructing certain models. For example, if stock prices are assumed to be normally
distributed, there is a positive (although perhaps tiny) probability that the stock price becomes
negative. This is impossible under a log-normal model.

LOGNORMAL DISTRIBUTION www.midhafin.com

• A log-normal distribution can be denoted as
𝑌 ∼ 𝐿𝑜𝑔𝑁 𝜇, 𝜎 , or equivalently as 𝑙𝑛 𝑌 ∼ 𝑁 𝜇, 𝜎
where
𝑙𝑛 𝑌 is normally distributed with mean 𝜇 and variance 𝜎 .

• Here, 𝑌 is a non-linear transformation of a normal random variable

• The mean of 𝑌 is
𝐸𝑌 𝑒

• The variance of 𝑌 is
𝑉𝑌 𝑒 1 𝑒

• Lognormal distribution is positively skewed.

• It peaks at 𝑒
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
47 of 65

LOGNORMAL DISTRIBUTION www.midhafin.com

• The PDF of a log-normal is given by: • The CDF of 𝑌 is
1 𝑛𝑦 𝜇
𝑓 𝑦 𝑒 𝐹 𝑦 Φ
𝑦 2𝜋𝜎 𝜎

CHI SQUARE DISTRIBUTION www.midhafin.com

• The 𝜒 (chi-squared) distribution is frequently encountered when testing hypotheses about
model parameters. It is also used when modeling variables that are always positive, (e.g., the VIX
Index).

• A 𝜒 random variable is defined as the sum of the squares of 𝜈 (Greek nu) independent
standard normal random variables

𝑌 𝑍

• Note that a 𝜒 distribution has 𝜈 degrees of freedom, a concept that arises when dealing with
models that have 𝑘 parameters using 𝑛 data points. Degrees of freedom measure the amount of
data that is available to test model parameters, because estimating model parameters requires a
minimum number of observations (e.g., 𝑘). In many models, the degree of freedom used in
testing is 𝑛 𝑘.

CHI SQUARE DISTRIBUTION www.midhafin.com

• The mean of 𝑌~𝜒 is
𝐸𝑌 𝜈

• The variance of 𝑌 is
𝑉𝑌 2𝜈

CHI SQUARE DISTRIBUTION www.midhafin.com

The PDF of a 𝜒 random variable is • The CDF of 𝑌 is
1 𝜈 𝜈 𝑦
𝑓 𝑦 𝐹 𝑦 𝑃 ,
𝜈 𝜈 𝑦 2 1 𝑒 2 2
2 Γ where 𝑃 is the regularized gamma function.
2 2
where Γ 𝑥 is known as the Gamma function.

STUDENT’S T DISTRIBUTION www.midhafin.com

• The Student’s 𝑡 distribution is closely related to the normal, but it has heavier tails. It was
originally developed for testing hypotheses using small samples.

• A Student’s 𝑡 is a one-parameter distribution. This parameter, denoted by 𝜈, is also called the

degrees of freedom parameter. While it affects many aspects of the distribution, the most
important effect is on the shape of the tails of the distribution.

The Student’s 𝑡 is the distribution of

𝑍
𝑌
𝑊
𝜈
• where
𝑍 is a standard normal
𝑊 is a 𝜒 random variable, and
𝑍 and 𝑊 are independent.

STUDENT’S T DISTRIBUTION www.midhafin.com

• Dividing a standard normal by another random variable produces heavier tails than the standard
normal. This is true for all values of 𝜈, although a Student’s t converges to a standard normal as
𝜈 → ∞.

• If 𝑌 ∼ 𝑡 , then the mean is

𝐸𝑌 0
The mean is only finite if 𝜈 1

• The variance of 𝑌 is
𝜈
𝑉𝑌
𝜈 2
The variance is only finite if 𝜈 2

• The kurtosis of 𝑌 is
𝜈 2
𝑘𝑢𝑟𝑡𝑜𝑠𝑖𝑠 𝑌 3
𝜈 4
The kurtosis is defined for 𝜈 4 and is always larger than 3

STUDENT’S T DISTRIBUTION www.midhafin.com

• Using the basic result that
𝑉 𝑎𝑌 𝑎 𝑉𝑌
It is easy to see that
𝜈 2
𝑉 𝑌 1
𝜈
when 𝑌 ∼ 𝑡 .

This distribution is known as a standardized Student’s 𝑡, because it has mean 0 and variance 1
for any value of 𝜈.

• The above distribution can be rescaled to have any variance and re-centered to have any mean if
𝜈 2. For example, 𝑋 𝜇 𝜎𝑆 is known as a generalized Student’s 𝑡 and has mean 𝜇,
variance 𝜎 , and degrees of freedom parameter 𝜈. It is parameterized with three parameters
reflecting the mean, variance, and degrees of freedom, and is denoted as
𝐺𝑒𝑛. 𝑡 𝜇, 𝜎

STUDENT’S T DISTRIBUTION www.midhafin.com

• As compared to the normal distribution, the generalized 𝑡 is better suited to model the returns of
many assets. It captures the heavy tails in asset returns while retaining the flexibility of a normal
random variable to directly set the mean and variance.

F DISTRIBUTION www.midhafin.com
• The 𝐹 is another distribution that is commonly encountered when testing hypotheses about
model parameters. The 𝐹 has two parameters, 𝜈 and 𝜈 , respectively known as the numerator
and denominator degrees of freedom.

• An 𝐹 distribution is defined as the ratio of two independent 𝜒 random variables where each has
been divided by its degree of freedom
𝑊
𝜈
𝑌 𝑊
𝜈
where 𝑊 ∼ 𝜒 , and 𝑊 ∼ 𝜒 , and
𝑊 and 𝑊 are independent.

F DISTRIBUTION www.midhafin.com
• If 𝑌 ∼ 𝐹 , , then the mean of 𝑌 is
𝜈
𝐸𝑌
𝜈 2
which is only finite when 𝜈 is larger than 2.

• The variance of 𝑌 is
2𝜈 𝜈 𝜈 2
𝑉𝑌
𝜈 𝜈 2 𝜈 4

and is only finite for 𝜈 4.

F DISTRIBUTION www.midhafin.com
• When using an 𝐹 in hypothesis testing, 𝜈 is usually determined by the hypothesis being tested
and is typically small (e.g., 1, 2, 3, …), while 𝜈 is related to the sample size (and so is relatively
large).

EXPONENTIAL DISTRIBUTION www.midhafin.com

• The exponential distribution uses a single parameter, 𝛽, that determines both the mean and
variance.

• If 𝑌 ∼ 𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙 𝛽 , then mean of 𝑌 is

𝐸𝑌 𝛽
• The variance of 𝑌 is
𝑉𝑌 𝛽

• The exponential distribution is closely related to the Poisson distribution.

For example, suppose 𝑋 is a random variable that measures the number of loan defaults per
quarter.
If 𝑋 is Poisson distributed with parameter 𝛽, then the time between each subsequent loan
default has as an exponential distribution with parameter 𝛽.

• Exponential variables are also memoryless, meaning that their distributions are independent
of their histories.
____________________________________________________________________________________________________________
www.midhafin.com
© 2019 MidhaFin
REPRODUCTION OF ANY PART OF THIS DOCUMENT IS STRICTLY PROHIBITED
59 of 65

EXPONENTIAL DISTRIBUTION www.midhafin.com

• The PDF of an 𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙 𝛽 is • The CDF is
𝑦
1 𝑦 𝐹 𝑦 1 exp
𝑓 𝑦 exp ,𝑦 0 𝛽
𝛽 𝛽

EXPONENTIAL DISTRIBUTION - EXAMPLE www.midhafin.com

• Assume that the time to default for a specific segment of credit card consumers is exponentially
distributed with a 𝛽 of 5 years. Find the probability that the customer will not default before
year 6

BETA DISTRIBUTION www.midhafin.com

• The beta distribution applies to continuous random variables with outcomes between 0 and 1.
It is commonly used to model probabilities that naturally fall into this range. The beta
distribution has two parameters, 𝛼 and 𝛽, that jointly determine the mean and variance of a
beta-distributed random variable.

• If 𝑌 ∼ 𝐵𝑒𝑡𝑎 𝛼, 𝛽 , then mean of 𝑌 is

𝛼
𝐸𝑌
𝛼 𝛽
• The variance of 𝑌 is
𝛼𝛽
𝑉𝑌
𝛼 𝛽 𝛼 𝛽 1

BETA DISTRIBUTION www.midhafin.com

The PDF of a beta 𝛼, 𝛽 is • The CDF is
𝑦 1 𝑦 𝐹 𝑦 𝐼 , 𝛼, 𝛽
𝑓 𝑦
𝐵 𝛼, 𝛽 where 𝐼 , 𝛼, 𝛽 is the regularized incomplete
beta function.
where 𝐵 𝛼, 𝛽 , and
Γ ⋅ is the Gamma function.

MIXTURES OF DISTRIBUTIONS www.midhafin.com

• Mixture distributions build new, complex distributions using two or more component
distributions.

• A two-component mixture first draws a value from a Bernoulli random variable. Then,
depending on the value (0 or 1), draws from one of two component distributions. This
structure makes it simple to compute the CDF of the mixture when the components are
normal random variables

MIXTURES OF DISTRIBUTIONS www.midhafin.com

• Mixing components with different means and variances produces a distribution that is both
skewed and heavy-tailed.

END OF LECTURE

THANK YOU

MATHEMATICS Lesson Note JSS 3
100% (2)
MATHEMATICS Lesson Note JSS 3
22 pages
Math & Eng Eco General Evaluation Exam Problems and Elements (With Answers and Solutions)
100% (1)
Math & Eng Eco General Evaluation Exam Problems and Elements (With Answers and Solutions)
53 pages
Currency Futures and Options - Final
No ratings yet
Currency Futures and Options - Final
107 pages
QT-Random Variable and Probability Distribution-1
No ratings yet
QT-Random Variable and Probability Distribution-1
4 pages
Chapter 9 Financial Instruments: Learning Objectives
No ratings yet
Chapter 9 Financial Instruments: Learning Objectives
41 pages
STAT2120: Categorical Data Analysis Chapter 1: Introduction
No ratings yet
STAT2120: Categorical Data Analysis Chapter 1: Introduction
51 pages
Hull OFOD10e MultipleChoice Questions Only Ch21
No ratings yet
Hull OFOD10e MultipleChoice Questions Only Ch21
4 pages
Notes Dvi
No ratings yet
Notes Dvi
34 pages
SI Chapter-1
No ratings yet
SI Chapter-1
30 pages
Statistics For Business and Economics: Discrete Random Variables and Probability Distributions
No ratings yet
Statistics For Business and Economics: Discrete Random Variables and Probability Distributions
82 pages
Stats
No ratings yet
Stats
24 pages
QTA 2 - Random Variables
No ratings yet
QTA 2 - Random Variables
50 pages
Study Note Chap 3
No ratings yet
Study Note Chap 3
32 pages
Lec 01
No ratings yet
Lec 01
44 pages
Topic Two. Random Variable and Probability Distribution
No ratings yet
Topic Two. Random Variable and Probability Distribution
43 pages
Group 5
No ratings yet
Group 5
53 pages
STT201
No ratings yet
STT201
19 pages
Related Rates and Optimization
No ratings yet
Related Rates and Optimization
18 pages
Notes On Unit 3
No ratings yet
Notes On Unit 3
42 pages
FRM Part 1: Book 2 - Quantitative Analysis
No ratings yet
FRM Part 1: Book 2 - Quantitative Analysis
24 pages
Slide FU - W3
No ratings yet
Slide FU - W3
34 pages
Lecture Statistics 07
No ratings yet
Lecture Statistics 07
26 pages
Akhilesh Awasthy - 3 - Functioning of Power Exchanges
No ratings yet
Akhilesh Awasthy - 3 - Functioning of Power Exchanges
57 pages
Day 02-Random Variable and Probability - Part (I)
No ratings yet
Day 02-Random Variable and Probability - Part (I)
34 pages
Msiii PDF
No ratings yet
Msiii PDF
118 pages
2 - Basics of Financial Statistics
No ratings yet
2 - Basics of Financial Statistics
32 pages
Random Variables Distributions
No ratings yet
Random Variables Distributions
36 pages
Random Variables
No ratings yet
Random Variables
31 pages
Prob. Distri.
No ratings yet
Prob. Distri.
36 pages
Theoretical Distributions: Bernoulli's Distribution Binomial Distributions
No ratings yet
Theoretical Distributions: Bernoulli's Distribution Binomial Distributions
35 pages
ExecCourse1 9-20pdf
No ratings yet
ExecCourse1 9-20pdf
44 pages
Lecture 3
No ratings yet
Lecture 3
39 pages
Mstat Note7 Random Variable f23
No ratings yet
Mstat Note7 Random Variable f23
76 pages
Theoretical Physics: Course Codes: Phys2325/Phys3150
No ratings yet
Theoretical Physics: Course Codes: Phys2325/Phys3150
36 pages
Lecture Note 3
No ratings yet
Lecture Note 3
11 pages
Calculus Made Ridiculusly Easy
92% (12)
Calculus Made Ridiculusly Easy
250 pages
Probability 2 FPM
No ratings yet
Probability 2 FPM
55 pages
ST2187 - Block 6 Common Probability Distributions in Business Applications
No ratings yet
ST2187 - Block 6 Common Probability Distributions in Business Applications
15 pages
Discrete Random Variable
No ratings yet
Discrete Random Variable
42 pages
Bachelor Thesis
No ratings yet
Bachelor Thesis
82 pages
Tài liệu 5
No ratings yet
Tài liệu 5
19 pages
Important PMFs and PDFs
No ratings yet
Important PMFs and PDFs
7 pages
Lecture Slides - Chapter 3
No ratings yet
Lecture Slides - Chapter 3
20 pages
Chapter 2
No ratings yet
Chapter 2
8 pages
Discrete Probability Distribution
No ratings yet
Discrete Probability Distribution
18 pages
Topic 4: Some Special Distributions: Rohini Somanathan Course 003, 2014-2015
No ratings yet
Topic 4: Some Special Distributions: Rohini Somanathan Course 003, 2014-2015
31 pages
2 2+random+variables
No ratings yet
2 2+random+variables
23 pages
Tetrad Formulation of The Einstein Field Equations: The Newman-Penrose Equations
No ratings yet
Tetrad Formulation of The Einstein Field Equations: The Newman-Penrose Equations
16 pages
FRM Part 1: Distributions
No ratings yet
FRM Part 1: Distributions
25 pages
What Is A Probability Distribution
No ratings yet
What Is A Probability Distribution
14 pages
Business Inferential Statistics Lessons
No ratings yet
Business Inferential Statistics Lessons
7 pages
A3 - Random Variables and Distributions
100% (1)
A3 - Random Variables and Distributions
19 pages
Section 4
No ratings yet
Section 4
26 pages
Chapter 3 Common Univariate Random Variables
No ratings yet
Chapter 3 Common Univariate Random Variables
7 pages
Chapter Three
No ratings yet
Chapter Three
12 pages
Inside Job Vocabulary
50% (2)
Inside Job Vocabulary
2 pages
STA80006 Weeks7-12 PDF
No ratings yet
STA80006 Weeks7-12 PDF
29 pages
Pricing and Hedging Spread Options in A Log-Normal Model: Bstract
No ratings yet
Pricing and Hedging Spread Options in A Log-Normal Model: Bstract
30 pages
Chapter 6
No ratings yet
Chapter 6
5 pages
Probability Distribution: Shreya Kanwar (16eemme023)
No ratings yet
Probability Distribution: Shreya Kanwar (16eemme023)
51 pages
Random Variables. Discrete Random Variables Continuous Random Variables
No ratings yet
Random Variables. Discrete Random Variables Continuous Random Variables
6 pages
Chapter 3 Discrete Probability Distributions - 2
No ratings yet
Chapter 3 Discrete Probability Distributions - 2
15 pages
Bus Stat CHP 6&7
No ratings yet
Bus Stat CHP 6&7
7 pages
Statistics Master
No ratings yet
Statistics Master
12 pages
Common Probability Distributions: 1.1 Bernoulli Distribution
No ratings yet
Common Probability Distributions: 1.1 Bernoulli Distribution
6 pages
D Models
No ratings yet
D Models
5 pages
LNCom Distr Poiss Proc
No ratings yet
LNCom Distr Poiss Proc
5 pages
Ex No-4
No ratings yet
Ex No-4
4 pages
Discrete RV and Binomial, Poisson, and Hypergeometric Distributions - Lecture Notes (DR AKM Azad)
No ratings yet
Discrete RV and Binomial, Poisson, and Hypergeometric Distributions - Lecture Notes (DR AKM Azad)
3 pages
Viscoelasticity 05 Laplace
No ratings yet
Viscoelasticity 05 Laplace
7 pages
11.1 Discrete Probability Distributions
No ratings yet
11.1 Discrete Probability Distributions
5 pages
Banking Internal Risk Models1
No ratings yet
Banking Internal Risk Models1
12 pages
STAT 342 Statistical Methods For Engineers
No ratings yet
STAT 342 Statistical Methods For Engineers
26 pages
Barrier Option Pricing For Assets With Markov-Modulated Dividends
No ratings yet
Barrier Option Pricing For Assets With Markov-Modulated Dividends
15 pages
Kuliah - Solusi Numerik Persamaan Diferensial - MHS
No ratings yet
Kuliah - Solusi Numerik Persamaan Diferensial - MHS
29 pages
CH 10 Hull Fundamentals 8 The D
No ratings yet
CH 10 Hull Fundamentals 8 The D
20 pages
Engineering Math
No ratings yet
Engineering Math
29 pages
Commodity Deravative Intrrduction
100% (1)
Commodity Deravative Intrrduction
15 pages
MA 205 Complex Analysis: Shripad M. Garge. IIT Bombay
No ratings yet
MA 205 Complex Analysis: Shripad M. Garge. IIT Bombay
16 pages
Geert Booij - Inherent and Contextual Inflection
No ratings yet
Geert Booij - Inherent and Contextual Inflection
17 pages
Multipole Expansions in Two Dimensions
No ratings yet
Multipole Expansions in Two Dimensions
10 pages
Prime Broker Overview: The Ice Link Solution
No ratings yet
Prime Broker Overview: The Ice Link Solution
2 pages
Part Diff and Wave Eqs
No ratings yet
Part Diff and Wave Eqs
2 pages
Belgrade 2004 Yoy
No ratings yet
Belgrade 2004 Yoy
7 pages
Spanning Multi-Asset Payoffs With Relus
No ratings yet
Spanning Multi-Asset Payoffs With Relus
43 pages
FE Course Part1
No ratings yet
FE Course Part1
116 pages
Figlewski 2018
No ratings yet
Figlewski 2018
33 pages
Do-It-Yourself Financial Plans - VeriPlan User Guide - 2025
From Everand
Do-It-Yourself Financial Plans - VeriPlan User Guide - 2025
Lawrence J. Russell
No ratings yet
Can Hoover Dam’s Design Principles Help Us Solve the Retirement Income Problem?
From Everand
Can Hoover Dam’s Design Principles Help Us Solve the Retirement Income Problem?
Dale Maley
No ratings yet