0% found this document useful (0 votes)
67 views54 pages

CVE 303 - 3. Probability Distributions

Uploaded by

adnan.buny
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
0% found this document useful (0 votes)
67 views54 pages

CVE 303 - 3. Probability Distributions

Uploaded by

adnan.buny
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/ 54

CVE 303

Probability and Statistics for Civil Engineers


Assist. Prof. Dr. Hasan Zaifoğlu

Middle East Technical University Northern Cyprus Campus


Civil Engineering Program

3. Probability Distributions
Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Discrete Probability Distributions

Probability
Distributions

Discrete
Probability
Distributions

Binomial

Poisson
Binomial Probability Distribution
• Let us consider a fix number of observations, 𝑛, where only two
outcomes are possible.

• Suppose one of the outcomes corresponds to ‘success’ and the other


to ‘failure’ (Two mutually exclusive and collectively exhaustive
categories)

• The probability of success will be denoted by 𝑝, and the probability of


failure by 𝑞 = 1 − 𝑝.

– e.g., if the success in the throw a die is equated with the throw of a
six, then, 𝑝 = 1/6 and 𝑞 = 1 − 1Τ6 = 5/6.

• Constant probability for each observation

• Observations are independent


Binomial Probability Distribution
• For 𝑛 number of trials (or observations, or sample size), the probability
of 𝑥 successes can be computed by using the pmf of binomial
distribution:
𝒏 𝒙 𝒏−𝒙
𝑷 𝑿=𝒙 = 𝒑 𝒒
𝒙
𝑛 𝑛!
where = denotes the number of combinations of 𝑛 different
𝑥 𝑥! 𝑛−𝑥 !
things taken 𝑥 at a time. If X ~ B(𝑛, 𝑝), that is, 𝑋 is a binomially distributed
random variable, then, the parameters of the variable 𝑋 are:

– 𝑬 𝑿 = 𝒏𝒑

– 𝑽𝒂𝒓 𝑿 = 𝒏𝒑𝒒 = 𝒏𝒑(𝟏 − 𝒑)


Binomial Probability Distribution
The shape of the binomial distribution depends on the values of
𝑝 and 𝑛
Binomial Probability Distribution
Example: What is the probability of one success in five
observations if the probability of success is 0.1?

Answer to the Example:


𝑛 𝑥 𝑛−𝑥
𝑥=1 𝑃 𝑋=𝑥 = 𝑝 𝑞
𝑥
𝑛=5
𝑝 = 0.1 𝑃 𝑋=1 =
5
0.11 (1 − 0.1)5−1
1

= 5 × 0.1 × 0.94 = 𝟎. 𝟑𝟐𝟖𝟎𝟓

𝐸 𝑋 = 𝜇 = 𝑛𝑝 𝑉𝑎𝑟 𝑋 = 𝜎 2 = 𝑛𝑝𝑞
𝜇 = 5 × 0.1 = 0.5 𝜎 2 = 5 × 0.1 × 0.9 = 0.45
𝜎 = 0.6708
Binomial Probability Distribution
Example: The probability of successful bid for a
contractor is assumed to be 1/3. Since the variable 𝑋
denoted the number of successes in 3 bids,
a) Compute the probability of 0, 1, 2, and 3 successes
in 3 bids.

b) The probability that the contractor is successful at


least once in 3 bids.

c) The expected value of successes in 3 bids.


Binomial Probability Distribution
Answer to the Example:
𝑛 𝑥 𝑛−𝑥
𝑃 𝑋=𝑥 = 𝑝 𝑞
a) 𝑥
3
𝑃 𝑋=0 = (1/3)0 (2/3)3 = 8/27
0
3
𝑃 𝑋=1 = (1/3)1 (2/3)2 = 12/27
1
3
𝑃 𝑋=2 = (1/3)2 (2/3)1 = 6/27
2
3
𝑃 𝑋=3 = (1/3)3 (2/3)0 = 1/27
3
Binomial Probability Distribution
Answer to the Example:

b)
𝑃 𝑋 ≥1 =1−𝑃 𝑋 <1
= 1 − 𝑃(𝑋 = 0)
8 19
=1− =
27 27
c)
𝐸 𝑋 = 𝜇 = 𝑛𝑝
1
=3× =1
3
Discrete Probability Distributions

Probability
Distributions

Discrete
Probability
Distributions

Binomial

Poisson
Poisson Distribution
The Poisson distribution is a discrete distribution that measures the
probability of a given number of events happening in a specified time
period.

• A discrete random variable 𝑋 is said to have a Poisson distribution and


it has a probability mass function (pmf) given by

𝒆−𝝀 𝝀𝒙
𝑷 𝑿 =
𝒙!
where 𝑥 is the number of events (or successes), 𝝀 (lambda) is the average
(expected) number of events, and 𝑒 is Euler's number (𝑒 = 2.71828 …).
Poisson Distribution
For the Poisson pmf given by
𝑒 −𝜆 𝜆𝑥
𝑃 𝑋 =
𝑥!
for 𝑥 = 0, 1, 2, 3, … and 𝜆 > 0.

The parameters of the variable 𝑋 are:

–𝑬𝑿 =𝝁=𝝀

– 𝑽𝒂𝒓 𝑿 = 𝝈𝟐 = 𝝀

–𝝈= 𝝀
Poisson Distribution
Assumptions and validity:

• 𝑥 is the number of times an event occurs in an interval


and 𝑥 can take values 0, 1, 2, ….
• The occurrence of one event does not affect the probability
that a second event will occur. That is, events occur
independently.
• The average rate at which events occur is independent of
any occurrences. For simplicity, this is usually assumed to be
constant, but may in practice vary with time.
• Two events cannot occur at exactly the same instant;
instead, at each very small sub-interval, either exactly one
event occurs, or no event occurs.
Poisson Distribution
The shape of the Poisson Distribution depends on the
parameter 𝜆

Poisson pmf, for 𝜆 = 0.2

Poisson pmf, for 𝜆 = 5


Poisson Distribution
Example: On a particular river, overflow floods occur
once every 100 years on average. By assuming the
Poisson model is appropriate,
Calculate the probability of 𝑥 = 0, 1, 2 overflow floods in
a 100-year interval.
Poisson Distribution
Answer to the Example:

Because the average event rate is one overflow flood


per 100 years, 𝜆 = 1
𝑒 −𝜆 𝜆𝑥 𝑒 −1 1𝑥
𝑃 𝑥 overflow floods in 100 years = =
𝑥! 𝑥!
𝑒 −1 10 𝑒 −1
𝑃 0 overflow floods in 100 years = = ≈ 0.368
0! 1
𝑒 −1 11 𝑒 −1
𝑃 1 overflow floods in 100 years = = ≈ 0.368
1! 1
𝑒 −1 12 𝑒 −1
𝑃 2 overflow floods in 100 years = = ≈ 0.184
2! 2
Continuous Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Continuous Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Uniform Distribution
• The uniform distribution is a probability distribution that has equal
probabilities for all possible outcomes of the random variable

• Also called a rectangular distribution

As implied by the name, the pdf is constant over a given interval, for
example, from 𝑎 to 𝑏, where 𝑎 < 𝑏, and takes the form

𝟏
𝒇 𝒙 = , for 𝑎 ≤ 𝑥 ≤ 𝑏
𝒃−𝒂
= 𝟎, otherwise

**All values of the variate between the lower limit 𝑎 and the upper limit 𝑏
are equally frequent or equally likely to occur.
Uniform Distribution
• The mean of a uniform distribution is
𝒂+𝒃
𝝁=
𝟐
• The standard deviation is

𝒃−𝒂 𝟐
𝝈=
𝟏𝟐
Uniform Distribution
Example: Find the pdf, the mean and the standard
deviation of the uniform probability distribution over the
range 2 ≤ 𝑋 ≤ 6.
Answer to the Example:
𝑓(𝑥)
1
0.25 𝑓 𝑥 = = 𝟎. 𝟐𝟓
6−2

2 6 𝑥

𝑎 +𝑏 2+6 𝑏−𝑎 2 6−2 2


𝜇= = =4 𝜎= = = 1.1547
2 2 12 12
The Uniform Distribution
Example: Use the same uniform probability distribution
as in the previous example and find 𝑃 3 ≤ 𝑥 ≤ 5 .
Answer to the Example:
𝑃 3 ≤ 𝑥 ≤ 5 = Base Height = 𝟐 𝟎. 𝟐𝟓 = 𝟎. 𝟓

𝑓(𝑥)

0.25

2 3 4 5 6 𝑥
Continuous Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Exponential Distribution
• The exponential distribution often used to model the length of time
between two occurrences of an event (the time between arrivals).

• We note that, the distribution is applicable for other variables such as


length and space in addition to time.

The probability density function (pdf) of an exponential distribution is

𝒇 𝒙 = 𝝀𝒆−𝝀𝒙 𝐟𝐨𝐫 𝒙 ≥ 𝟎, 𝝀 > 𝟎


=𝟎 𝐨𝐭𝐡𝐞𝐫𝐰𝐢𝐬𝐞
Using this result and considering the time 𝑇 between occurrences as the
random variable, we find the cdf of variable 𝑇 is

𝑭 𝑿 = 𝑻 ≤ 𝒙 = 𝒕 = 𝟏 − 𝒆−𝝀𝒙
Exponential Distribution
• The mean or expected value of an exponentially
distribution
𝟏 Mean Recurrence
𝑬𝑿 =
𝝀 Interval

• The variance is
𝟏
𝑽𝒂𝒓(𝑿) = 𝟐
𝝀
𝝀 is the rate at which events occur (often called the rate parameter).
It is the parameter of the distribution
Exponential Distribution
pdf cdf
Exponential Distribution
Example: Customers arrive at the service counter at the rate of
15 per hour. What is the probability that the arrival time
between consecutive customers is less than three minutes?

Answer to the Example:


• The mean number of arrivals per hour is 15, so 𝝀 = 𝟏𝟓
• Three minutes is 0.05 hours
• 𝑃 arrival time < 0.05 = 1 − 𝑒 −𝜆𝑥
• 𝑃 arrival time < 0.05 = 1 − 𝑒 − 15 0.05 = 0.5276
So there is a 52.76% probability that the arrival time
between successive customers is less than three minutes
Continuous Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Normal Distribution
• Bell Shaped’
• Symmetrical
• Mean, Median and Mode are Equal

• Location is determined by the mean, 𝝁


• Spread is determined by the standard deviation, 𝝈
• The random variable has an infinite theoretical range:
−∞<𝒙<∞
Normal Distribution

(Changing 𝝁 shifts the distribution left or right)

(Changing 𝝈 increases or decreases the spread)


Normal Distribution
• A large number of random variables encountered in practical
applications fit to the normal (Gaussian) distribution with the following
probability density function (pdf):

𝟏 𝒙−𝝁𝑿 𝟐
𝟏 −
𝟐 𝝈𝑿
𝒇 𝒙 = 𝒆
𝝈 𝟐𝝅
where −∞ < 𝒙 < ∞ and specified by two parameters, the mean 𝜇
(location parameter) and the standard deviation 𝜎 (scale parameter) of the
population.

• The notation 𝑵(𝝁, 𝝈𝟐 ) is used to indicate such a distribution.

• The parameters should be in the units of the variables, such as 𝒎𝟑 /𝒔


for the flow in a river used in hydrology.
Normal Distribution
• The analytical form of the cumilative distribution function (cdf) 𝐹(𝑥) of
the normal distribution cannot be obtained by numerical methods
(integration).

• In practice the cdf is tabulated numerically. A single table for the normal
distribution is prepared by standardizing the random variable as
follows:

(𝑿 − 𝝁𝑿 )
𝒁=
𝝈𝑿
where the standard normal variable 𝑍 has the mean 0 and standard
deviation 1.

• The distribution 𝑵(𝟎, 𝟏) of the variable 𝑍 is called the standard normal


distribution.
Normal Distribution
• The cdf of 𝑋 (or can be given for 𝑍) can be written as
𝒙−𝝁
𝑭 𝒛 =𝑷 𝑿≤𝒙 =𝑷 𝒁≤ = 𝑷(𝒁 ≤ 𝒛)
𝝈
• The Cumulative Standardized Normal table gives numerical solutions

• The Cumulative Standardized Normal table gives the probability less


than a desired value for 𝑍 (i.e., from negative infinity to 𝑍)

• Find the 𝑍 value for the known probability, then convert to 𝑋 units using
the formula:

𝒙 = 𝝁 + 𝒛𝝈
Cumulative Standardized Normal Table

Cumulative
Standardized
Normal table
Normal Distribution
How to read a value: Cumulative Standardized Normal table
𝑷 𝒁 ≤ 𝟏. 𝟗𝟔 =? The column gives the value of
𝑍 to the second decimal point
𝑍 0.00 0.01 0.02 …0.06

The row shows 0.0


The value within the
the value of 𝑍 0.1 table gives the
.
to the first . probability from Z =  
decimal point . up to the desired Z
1.9 0.975 value

2.0 0.975

𝑃 𝑍 ≤ 1.96 = 𝟎. 𝟗𝟕𝟓
0 1.96 Z
Normal Distribution
General Procedure for Finding Probabilities
To find 𝑷(𝒂 < 𝑿 < 𝒃) when 𝑿 is distributed normally:

• Draw the normal curve for the problem in terms of 𝑋

• Translate 𝑋-values to 𝑍-values

• Use the Standardized Normal Table


Normal Distribution
• 𝑍 is an 𝑁(0, 1) variate with pdf
𝟏 𝒛𝟐

𝒇 𝒛 = 𝒆 𝟐
𝝈 𝟐𝝅

• Note that the distribution is


the same, only the scale has
changed.
• We can express the problem
in original units ( 𝑋 ) or in
standardized units (𝑍)
Properties of the Normal Distribution
• 𝑋 is an 𝑁(𝜇, 𝜎 2 ) variate with pdf
𝟏 𝒙−𝝁𝑿 𝟐
𝟏 −
𝟐 𝝈𝑿
𝒇 𝒙 = 𝒆
𝝈 𝟐𝝅
𝒇(𝒙) 𝑃(a ≤ 𝑿 ≤ b )
= 𝑃( a < 𝑿 < b )
(Note that the
probability of any
individual value is zero)

a b 𝑿

**Probability is measured by the area under the curve


Properties of the Normal Distribution

There are some general rules:


• About 68% of values drawn from a normal distribution are
within one standard deviation 𝜎 away from the mean

• About 95% of the values lie within two standard deviations

• About 99.7% are within three standard deviations


Properties of the Normal Distribution
• Linear combinations of normal random variables are
also normally distributed:

– A linear transformation 𝑌 = 𝑎 + 𝑏𝑋 of an 𝑁(𝜇, 𝜎 2 ) random


variable 𝑋 makes 𝑌 an 𝑁(𝑎 + 𝑏𝜇, 𝑏 2 𝜎 2 ) random variable.

• The sum of two independent Normal random variables is


also normally distributed:

– If 𝑋1 ~𝑁(𝜇1 , 𝜎12 ) and 𝑋2 ~𝑁(𝜇2 , 𝜎22 ) are independent


random variables, then
𝑌 = 𝑋1 + 𝑋2 ~𝑁(𝜇1 + 𝜇2 , 𝜎12 + 𝜎22 )
Normal Distribution
Example: Annual depth of precipitation 𝑋 (cm) is
assumed to be normally distributed with 50 cm
mean and 10 cm standard deviation.
What is the probability of the precipitation to
remain in the range of 45 and 62 cm?

solved in the class


Normal Distribution
Answer to the Example:

𝑋: Annual Depth of Precipitation


• 𝜇𝑥 = 50 cm
X: N(50,10) 𝑃 45 ≤ 𝑋 ≤ 62 =?
• 𝜎𝑥 = 10 cm

The values of the standard normal variable (𝑧) corresponding to 45 to


62 are computed:

45 − 50 62 − 50
𝑧1 = = −0.5 , 𝑧2 = = 1.2
10 10
Normal Distribution
Answer to the Example:
From Cumulative Standardized Normal Table:
𝐹(−0.5) = 0.3085
𝐹(1.2) = 0.8849

𝑃 45 ≤ 𝑋 ≤ 62 = P −0.5 ≤ 𝑍 ≤ 1.2
= 𝐹 1.2 − 𝐹 −0.5
= 0.8849 − 0.3085
= 0.5764
Normal Distribution
Example: The load of a footing consists of the sum of the dead
load and moving load. These loads are assumed to be random
variables. The dead load 𝑋 has the mean 100 kN, standard 10
kN. The moving load 𝑌 has the mean 40 kN, standard deviation
10 kN.

If the design load corresponds to the risk


of exceedance of 5%, determine the
design load assuming the loads follow the
normal distribution.

solved in the class


Normal Distribution
Answer to the Example:
Given
The total load: 𝑈
• 𝜇𝑥 = 100 kN
• 𝜎𝑥 = 10 kN  𝑈 = 𝑋 + 𝑌 (will be also normal)
 𝜇𝑢 = 𝜇𝑥 + 𝜇𝑦 = 100 + 40 = 140 𝑘𝑁
• 𝜇𝑦 = 40 kN
• 𝜎𝑦 = 10 kN  𝜎𝑢 = 𝜎𝑥2 + 𝜎𝑦2 = 102 + 102 = 14.1 𝑘𝑁

 From cumulative standardized Normal table, the normal variable


variable corresponding to the exceedance probability of 0.05:
 Nonexceedence probability 1-0.05=0.95 𝒛 = 𝟏. 𝟔𝟓
The design load is:

𝑈𝑑 = 𝜇𝑢 + 𝑧𝜎𝑢 = 140 + 1.65 × 14.1 = 𝟏𝟔𝟑 kN


Continuous Probability Distributions

Probability
Distributions

Discrete Continuous
Probability Probability
Distributions Distributions

Binomial Uniform

Poisson Exponential

Normal
Lognormal
Lognormal Distribution
• It is often attempted to transform a nonnormal random variable to a
normal variable because the normal distribution has well known
properties and is easy to use. The most commonly used transformation
is the logarithmic transformation.
If the transformed variable
𝒀 = 𝐥𝐧𝑿 and 𝒀~𝑵(𝝁𝒀 , 𝝈𝟐𝒀 )
• Fits the normal distribution, then the distribution of the origional variable
𝑋 is called lognormal:
𝟐
𝟏 𝐥𝐧(𝒙)−𝝁𝒀
𝟏 −
𝟐 𝝈𝒀
𝒇 𝒙 = 𝒆
𝒙𝝈𝒀 𝟐𝝅
where 0 ≤ 𝒙 < ∞.
Lognormal Distribution
The lognormal pdf for two different pairs of values of the
parameters:

𝐘 = 𝐥𝐧 𝑿
𝝁𝒀 = 𝝁𝐥𝐧(𝑿)
𝝈𝒀 = 𝝈𝐥𝐧(𝑿)
Lognormal Distribution
• The standardized value, 𝑍

𝐥𝐧(𝑿) − 𝝁𝒀
𝒁=
𝝈𝒀
• The lognormal distribution parameters (𝜇𝑌 and 𝜎𝑌 ) are related to the
parameters of 𝑋 (𝜇𝑋 and 𝜎𝑋 ):
𝝁𝒙 𝟎.𝟓
𝝁𝒀 = 𝐥𝐧 𝟎.𝟓
𝝈𝟐𝑿
𝝈𝟐𝑿 𝝈𝒀 = 𝐥𝐧 +𝟏
+𝟏 𝝁𝟐𝑿
𝝁𝟐𝑿

𝝈𝟐𝒀 𝝈𝟐𝒀
𝟎.𝟓
𝝁𝒀 + 𝟐 𝝈𝑿 = 𝝁𝑿 𝒆 −𝟏
𝝁𝑿 = 𝒆
Lognormal Distribution
Example: Annual depth of precipitation 𝑋 (cm)
with 50 cm mean and 10 cm standard deviation is
assumed to be lognormally distributed.
What is the probability of the precipitation to
remain in the range of 45 and 62 cm?

solved in the class


Lognormal Distribution
Answer to the Example:
• 𝜇𝑥 = 50 cm
𝑋 ∶ 𝐿𝑁 (𝜇𝑌 , 𝜎𝑌2 )
• 𝜎𝑥 = 10 cm

 The parameters of the variable 𝑌 = ln𝑋:

50
 𝜇𝑌 = 𝑙𝑛 0.5 = 3.892 X: LN (3.892, 0.04)
102
+1
502
0.5
102
 𝜎𝑌 = 𝑙𝑛 +1 = 0.198
502

𝜎𝑌2 = 0.04
Lognormal Distribution
Answer to the Example:
𝑌 values corresponding to 𝑋1 = 45 and 𝑋2 = 62 :
 𝑌1 = ln 45 = 3.087
 𝑌2 = ln 62 = 4.127
𝑙𝑛𝑋−𝜇𝑌
 𝑧=
𝜎𝑌
3.087−3.892
 𝑧1 = = −0.431
0.198
4.127−3.892
 𝑧2 = = 1.187
0.198

𝑃 45 ≤ 𝑋 ≤ 62 = 𝑃 3.807 ≤ 𝑌 ≤ 4.127 = 𝑃 −0.431 ≤ 𝑍 ≤ 1.187 =?

= 𝐹 1.187 − 𝐹 −0.431 = 0.8824 − 0.332


= 0.5492
= 54.92%
End of Topic 3

You might also like