Scribd
Upload
20 views29 pages

Lecture 7. Typical Probability Distribution For Continuous RV

This lecture focuses on probability distribution functions for continuous random variables, highlighting common distributions such as normal, log-normal, and exponential. It explains the properties and applications of these distributions, including their parameters and sample spaces. Additionally, it introduces the beta distribution for modeling random variables limited to finite intervals.

Uploaded by

tszhonchan5
Copyright
© © All Rights Reserved
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
20 views29 pages

Lecture 7. Typical Probability Distribution For Continuous RV

This lecture focuses on probability distribution functions for continuous random variables, highlighting common distributions such as normal, log-normal, and exponential. It explains the properties and applications of these distributions, including their parameters and sample spaces. Additionally, it introduces the beta distribution for modeling random variables limited to finite intervals.

Uploaded by

tszhonchan5
Copyright
© © All Rights Reserved
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/ 29

CIVL2160

Modelling Systems with


Uncertainties
Lecture 7
Probability distribution I
For continuous RV
Zhe (Walter) Wang
Department of Civil & Environmental Engineering
Probability distribution function

• Any functions that pass the sanity check could be a legitimate


distribution (PMF for discrete and PDF for continuous) function for a
random variable
∞
෍ 𝑝𝑋 𝑥𝑖 = 1 න 𝑓𝑋 (𝑥) 𝑑𝑥 = 1
𝑎𝑙𝑙 𝑥𝑖 −∞

• However, there are a number of both discrete and continuous


distribution functions that are especially common and therefore
useful, because:
• The function is the result of an underlying physical process and can be derived
on the basis of certain physically reasonable assumptions
Probability distribution function

• Common continuous distributions


• Normal distribution, log-normal distribution, exponential distribution
• Common discrete distributions (for occurrence models)
• Bernoulli sequence
• Binomial distribution, geometric distribution and negative binomial distribution
• Poisson process
• Poisson, exponential, gamma
• This and next lectures will focus on the formulation and properties of
some of these common probability functions
01. Normal

Agenda 02. Log-normal

03. Exponential

04. Beta and others


Normal (Gaussian) distribution

• Important in the theory of probability


• Central limit theorem

• Prevalent in applications
• Convenient analytical properties
• Model of noise consisting of many, small independent noise terms
• Often used in the natural and social sciences to represent real-valued random
variables whose distributions are not known
Normal (Gaussian) distribution

• Parameter: 𝜇, 𝜎
• Sample space: 𝑥 ∈ ℝ
• Notation: 𝑋~𝒩(𝜇, 𝜎 2 )

Standard normal distribution General normal distribution


• 𝜇=0
• 𝜎=1

Z-score
Normal (Gaussian) distribution

• Parameter: 𝜇, 𝜎

Bell curve
Normal (Gaussian) distribution: properties

For normal random variable 𝑋~𝒩(𝜇, 𝜎 2 )


•E𝑋 =𝜇
• var 𝑋 = 𝜎 2

• Let 𝑌 = 𝑎𝑋 + 𝑏
• 𝑌 also follow normal distribution
• E 𝑌 = 𝑎𝜇 + 𝑏
• var 𝑌 = 𝑎2 𝜎 2
• 𝑌~𝒩(𝑎𝜇 + 𝑏, 𝑎2 𝜎 2 )
Standard normal table

• Motivation
• Both PDF and CDF of normal distribution are not easy to calculate

• Normal distribution is so widely used


• Solution: Standard normal table
• For standard normal distribution, i.e. 𝑋~𝒩 0,1
• How about general normal distribution?
• Convert to standard normal first
• Benefit: one table is enough!
Standard normal table

• This is for CDF, not for PDF!


𝛷 0 = 0.5
𝛷 1 = 0.8413
𝛷 −1 = 1 − 𝛷 1 = 0.1587
• Normal distribution is symmetric

P𝑋 0.23 < 𝑋 < 0.25 = 𝛷 0.25 − 𝛷 0.23


= 0.5987 − 0.5910 = 0.0077
For general normal distribution

• Express the event of interest in terms of


standard normal
• Assume 𝑋~𝒩 6,4
• Calculate P𝑋 2 < 𝑋 < 8

• Solution
𝑋−𝜇
• If 𝑍 = , then 𝑍~𝒩 0,1
𝜎
2−6 8−6
• P𝑋 2 < 𝑋 < 8 = P𝑍 <𝑍<
2 2
= P𝑍 −2 < 𝑍 < 1
= 𝛷 1 − 𝛷 −2 =
0.8413 − (1 − 0.9772)
Probability distribution

• A good video to learn more about normal distribution and for fun:
https://www.youtube.com/watch?v=cy8r7WSuT1I
01. Normal

Agenda 02. Log-normal

03. Exponential

04. Beta and others


Log-normal distribution: motivation

• Normal distribution
• The random variable could have both positive and negative values
• In many fields, the random variable could only have positive values
• The project completion time
• GDP of a country
• Etc.
• Log-normal distribution
• A convenient and useful model for measurements in exact and engineering
sciences, as well as medicine, economics and other topics (e.g., energies,
concentrations, lengths, prices of financial instruments, and other metrics).
Log-normal distribution: definition

• If the random variable 𝑋 is log-normally distributed, then 𝑌 =


ln(𝑋) has a normal distribution
• Or, if the random variable 𝑌 has a normal distribution, then 𝑋 =
exp(𝑌) has a log-normal distribution

• Parameter: 𝜇, 𝜎
• Sample space: 𝑥 ∈ ℝ+
• Notation: 𝑋~Lognormal(𝝁, 𝝈2 )
1 1 ln 𝑥 − 𝝁 2
𝑓𝑋 𝑥 = exp[− ( ) ] 𝑓𝑜𝑟 𝑥 ≥ 0
𝝈 2𝜋𝑥 2 𝝈
Log-normal vs. normal

• 𝑋~Lognormal(𝝁, 𝝈2 )
1 1 𝐥𝐧 𝑥 − 𝝁 2
𝑓𝑋 𝑥 = exp[− ( ) ] 𝑓𝑜𝑟 𝑥 ≥ 𝟎
𝝈 2𝜋𝒙 2 𝝈
Mean Standard deviation
𝝈𝟐 𝝈𝟐
𝐞𝐱𝐩 𝝁 + 𝐞𝐱𝐩 𝝈𝟐 − 𝟏𝐞𝐱𝐩(𝝁 + )
𝟐 𝟐
• 𝑋~𝒩(𝝁, 𝝈2 )
1 1 𝑥−𝝁 2
𝑓𝑋 𝑥 = exp[− ( ) ] 𝑓𝑜𝑟 −∞<𝑥 <∞
𝝈 2𝜋 2 𝝈
Mean Standard deviation
𝝁 𝝈
Log-normal distribution

• Parameter: 𝜇, 𝜎
1 1 ln 𝑥 − 𝜇 2
𝑓𝑋 𝑥 = exp[− ( ) ]
𝜎 2𝜋𝑥 2 𝜎
Example

• Suppose T is a r.v. about project completion time


• We know that T is normal with a mean, 𝜇, of 30 and 𝑃(𝑇<40)=0.9
• Question (a): What is 𝑷(𝑻<𝟓𝟎)?

• First of all, we do not know 𝝈


40 − 30
𝑃 𝑇 < 40 = 0.9 Φ = 0.9 𝜎 = 7.81
𝜎
• Then,
50 − 30
𝑃 𝑇 < 50 = Φ = Φ 2.56 = 0.9948
7.81
Example

• Question 2(b): What if T is lognormal, but with the same mean and
standard deviation
ln(50) − 𝜇
𝑃 𝑇 < 50 = 𝐹𝑋 50 = Φ
𝜎
• We know
𝜎2
• mean = exp 𝜇 + = 30 = 𝛾,
2
𝜎2
• standard deviation = exp 𝜎2 − 1exp 𝜇 + = 7.81 = 𝜆
2
• We need to back calculate 𝜇 and 𝜎
𝜆
exp 𝜎 2 − 1 = = 0.26
𝛾
ln(50) − 3.367
𝜎= ln 1 +0.262≈ 0.26 𝑃 𝑇 < 50 = 𝐹𝑋 50 = Φ = Φ 2.09 = 0.9817
0.26
1 2
𝜇 = ln 30 − 𝜎 = 3.367
2
01. Normal

Agenda 02. Log-normal

03. Exponential

04. Beta and others


Exponential distribution: motivation

• The exponential distribution or negative exponential distribution is


the probability distribution of the distance between events in a
Poisson process
• Poisson process: a process in which events occur continuously and
independently at a constant average rate
• We will learn Poisson process in greater details in Lecture 9
Exponential distribution: definition

• Parameter: 𝜆
• Sample space: 𝑥 ∈ ℝ+
• Notation:𝑋~Exp(𝜆)

𝑓𝑋 𝑥 = 𝜆 exp[−𝜆𝑥] 𝑓𝑜𝑟 𝑥 ≥ 0

Mean Standard Median Mode Skewness


deviation
1 1 ln2 0 2
𝜆 𝜆 𝜆
Exponential distribution

• Parameter: 𝜆

𝑓𝑋 𝑥 = 𝜆 exp[−𝜆𝑥] 𝑓𝑜𝑟 𝑥 ≥ 0 𝐹𝑋 𝑥 = 1 − exp[−𝜆𝑥] 𝑓𝑜𝑟 𝑥 ≥ 0


Example

• Question:
• Given that the random variable of earthquake magnitude follows the
Exponential distribution.
• The mean earthquake magnitude is 5 in Richter scale
• What is the probability of next quake larger than 7?
• First of all
1
𝐸 𝑋 = =5 𝜆 = 0.2
𝜆
• Then, ∞
∞
𝑃 𝑀>7 =න 0.2𝑒 −0.2𝑥 𝑑𝑥 = −𝑒 −0.2𝑥 | = 𝟎. 𝟐𝟓
7
7
01. Normal

Agenda 02. Log-normal

03. Exponential

04. Beta and others


Beta distribution: motivation

• Normal (ℝ), log-normal (ℝ+ ), and exponential (ℝ+ ) distributions are


all defined on an infinite range.
• However, many random events are defined on a limited intervals
• Percentage or proportions
• Location of the crack on a beam

• The beta distribution has been applied to model the behavior of


random variables limited to intervals of finite length in a wide variety
of disciplines.
• The beta distribution is a suitable model for the random behavior of
percentages and proportions.
Beta distribution: definition

Gamma function
• Parameter: 𝛼, 𝛽
• Sample space: 𝑥 ∈ [0,1]
• Notation:𝑋~Beta (𝛼, 𝛽)
Γ(𝛼 + 𝛽) 𝛼−1
𝑓𝑋 𝑥 = 𝑥 (1 − 𝑥)𝛽−1 for 𝟎 ≤ 𝒙 ≤ 𝟏
Γ(𝛼)Γ(𝛽)
Mean Standard deviation
𝛼
1 𝛼𝛽
𝛼+𝛽
𝛼+𝛽 𝛼+𝛽+1

Calculation of Beta distribution is not required. But you need to know why we need it.
Beta distribution

• Parameter: 𝜇, 𝜎
Γ(𝛼 + 𝛽) 𝛼−1
𝑓𝑋 𝑥 = 𝑥 (1 − 𝑥)𝛽−1 for 0 ≤ 𝑥 ≤ 1
Γ(𝛼)Γ(𝛽)
Other distributions

• Uniform distribution
• Triangular distribution
• Rayleigh distribution
•…

• Learn when you need

You might also like