Probability Distributions

Probability Distributions
CEE 201L. Uncertainty, Design, and Optimization

Department of Civil and Environmental Engineering
Duke University
Philip Scott Harvey, Henri P. Gavin and Jeffrey T. Scruggs
Spring 2022
1 Probability Distributions
Consider a continuous, random variable (rv) X with support over the domain X . The probability
density function (PDF) of X is the function fX (x) such that for any two numbers a and b in the
domain X , with a < b,
Z b
P [a < X ≤ b] = fX (x) dx
a
For fX (x) to be a proper distribution, it must satisfy the following two conditions:
1. The PDF fX (x) is positive-valued; fX (x) ≥ 0 for all values of x ∈ X .

R
2. The rule of total probability holds; the total area under fX (x) is 1; X fX (x) dx = 1.
Alternately, X may be described by its cumulative distribution function (CDF). The CDF
of X is the function FX (x) that gives, for any specified number x ∈ X , the probability that the
random variable X is less than or equal to the number x is written as P [X ≤ x]. For real values of
x, the CDF is defined by
Z b
FX (x) = P [X ≤ b] = fX (x) dx ,
−∞
so,
P [a < X ≤ b] = FX (b) − FX (a)
By the first fundamental theorem of calculus, the functions fX (x) and FX (x) are related as
d
fX (x) = FX (x)
dx
2 CEE 201L. Uncertainty, Design, and Optimization – Duke University – Spring 2022 – P.S.H., H.P.G. and J.T.S.
A few important characteristics of CDF’s of X are:
1. CDF’s, FX (x), are monotonic non-decreasing functions of x.

2. For any number a, P [X > a] = 1 − P [X ≤ a] = 1 − FX (a)
Rb
3. For any two numbers a and b with a < b, P [a < X ≤ b] = FX (b) − FX (a) = a fX (x)dx
2 Descriptors of random variables

The expected or mean value of a continuous random variable X with PDF fX (x) is the centroid
of the probability density. Z ∞
µX = E[X] = x fX (x) dx
−∞
The expected value of an arbitrary function of X, g(X), with respect to the PDF fX (x) is
Z ∞
µg(X) = E[g(X)] = g(x) fX (x) dx
−∞
The variance of a continuous rv X with PDF fX (x) and mean µX gives a quantitative measure of
how much spread or dispersion there is in the distribution of x values. The variance is calculated
as
Z ∞
2
σX = V[X] = (x − µX )2 fX (x) dx
−∞
=
=
=
=
p
The standard deviation (s.d.) of X is σX = V[X]. The coefficient of variation (c.o.v.) of
X is defined as the ratio of the standard deviation σX to the mean µX :

σX
cX =
µX
for non-zero mean. The c.o.v. is a normalized measure of dispersion (dimensionless).
A mode of a probability density function, fX (x), is a value of x such that the PDF is maximized;

d
fX (x) =0.
dx x=xmode
The median value, xm , is is the value of x such that

P [X ≤ xm ] = P [X > xm ] = FX (xm ) = 1 − FX (xm ) = 0.5 .
CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

Probability Distributions 3
3 Some common distributions
The National Institute of Standards and Technology (NIST) lists properties of nineteen commonly
used probability distributions in their Engineering Statistics Handbook. This section describes the
properties of seven distributions. For each of these distributions, this document provides figures
and equations for the PDF and CDF, equations for the mean and variance, the names of Matlab
functions to generate samples, and empirical distributions of such samples.
3.1 The Normal distribution
The Normal (or Gaussian) distribution is perhaps the most commonly used distribution function.
The notation X ∼ N (µX , σX 2 ) denotes that X is a normal random variable with mean µ
X and
2
variance σX . The standard normal random variable, Z, or “z-statistic”, is distributed as N (0, 1).
The probability density function of a standard normal random variable is so widely used it has its
own special symbol, φ(z), !
1 z2
φ(z) = √ exp −
2π 2
Any normally distributed random variable can be defined in terms of the standard normal random
variable, through the change of variables
X = µX + σX Z.
If X is normally distributed, it has the PDF

!
x − µX 1 (x − µX )2

fX (x) = φ =q exp − 2
σX 2
2πσX 2σX
There is no closed-form equation for the CDF of a normal random variable. Solving the integral
Z z
1 2 /2
Φ(z) = √ e−u du
2π −∞
would make you famous. Try it. The CDF of a normal random variable is expressed in terms of the
error function, erf(z). If X is normally distributed, P [X ≤ x] can be found from the standard
normal CDF
x − µX

P [X ≤ x] = FX (x) = Φ .
σX
Values for Φ(z) are tabulated and can be computed, e.g., the Matlab command . . .
Prob_X_le_x = normcdf(x,muX,sigX). The standard normal PDF is symmetric about z = 0,
so φ(−z) = φ(z), Φ(−z) = 1 − Φ(z), and P [X > x] = 1 − FX (x) = 1 − Φ ((x − µX )/σX ) =
Φ ((µX − x)/σX ).
The linear combination of two independent normal rv’s X1 and X2 (with means µ1 and µ2 and
variances σ12 and σ22 ) is also normally distributed,

aX1 + bX2 ∼ N aµ1 + bµ2 , a2 σ12 + b2 σ22 ,
and more specifically, aX − b ∼ N aµX − b, a2 σX

2 .


Given the probability of a normal rv, i.e., given P [X ≤ x], the associated value of x can be found
from the inverse standard normal CDF,
x − µX
= z = Φ−1 (P [X ≤ x]) .
σX
Values of the inverse standard normal CDF are tabulated, and can be computed, e.g., the Matlab
command . . . x = norminv(Prob_X_le_x,muX,sigX).
3.2 The Log-Normal distribution
The Normal distribution is symmetric and can be used to describe random variables that can take
positive as well as negative values, regardless of the value of the mean and standard deviation. For
many random quantities a negative value makes no sense (e.g., modulus of elasticity, air pressure,
and distance). Using a distribution which admits only positive values for such quantities eliminates
any possibility of non-sensical negative values. The log-normal distribution is such a distribution.
If ln X is normally distributed (i.e., ln X ∼ N (µln X , σln X )) then X is called a log-normal random

variable. In other words, if Y (= ln X) is normally distributed, eY (= X) is log-normally distributed.
P [Y ≤ y] FY (y)

Φ y−µ σY
Y
µY = µln X , σY2 = σln

2
X, P [ln X ≤ ln x] = Fln X (ln x) =
ln x−µln X

P [X ≤ x] FX (x) Φ σln X
The mean and standard deviation of a log-normal variable X are related to the mean and standard
deviation of ln X.
1 2 2

2

µln X = ln µX − σln σln X = ln 1 + (σX /µX )
2 X
If (σX /µX ) < 0.30, σln X ≈ (σX /µX ) = cX
The median, xm , is a useful parameter of log-normal rv’s. By definition of the median value, half
of the population lies above the median, and half lies below, so
ln xm − µln X

Φ = 0.5
σln X
ln xm − µln X
= Φ−1 (0.5) = 0
σln X
q
and, ln xm = µln X ↔ xm = exp(µln X ) ↔ µX = xm 1 + c2X
For the log-normal distribution xmode < xmedian < xmean . If cX < 0.15, xmedian ≈ xmean .
If ln X is normally distributed (X is log-normal) then (for cX < 0.3)

ln x − ln xm

P [X ≤ x] ≈ Φ
cX
2 ), and ln Y ∼ N (µ
If ln X ∼ N (µln X , σln 2 n m then
X ln Y , σln Y ), and Z = aX /Y
2
ln Z = ln a + n ln X − m ln Y ∼ N (µln Z , σln Z)
where µln Z = ln a + nµln X − mµln Y = ln a + n ln xm − m ln ym

2
σln 2 2 2 2 2 2 2
and Z = (nσln X ) + (mσln Y ) = n ln(1 + cX ) + m ln(1 + cY ) = ln(1 + cZ )

Uniform X ∼ U[a, b] Triangular X ∼ T (a, b, c)

a ≤ X ≤ b; a ≤ X ≤ b, a ≤ c ≤ b
1/(b-a) 2/(b-a)
p.d.f., f(x)
p.d.f., f(x)
0 0
a µ-σ µ µ+ σ b a µ-σ c µ µ+ σ µ+2 σ b
0.97
0.79 0.82
c.d.f., F(x)
c.d.f., F(x)
0.5 0.55
0.21 0.17
a µ-σ µ µ+ σ b a µ-σ c µ µ+ σ µ+2 σ b

x x

2(x−a)
 (b−a)(c−a) , x ∈ [a, c]
( 
1
x ∈ [a, b]

f (x) = b−a , f (x) = 2(b−x)
0, otherwise  (b−a)(b−c) , x ∈ [c, b]

 0, otherwise

  0, x≤a
 0, x≤a
 2
 (x−a) ,

x ∈ [a, c]
 
x−a (b−a)(c−a)
F (x) = b−a , x ∈ [a, b] F (x) = (b−x) 2

 1, x≥b  1 − (b−a)(b−c) , x ∈ [c, b]



1, x≥b

µX = 12 (a + b) µX = 31 (a + b + c)
2 = 1 1
σX 12 (b − a)2 2 =
σX 18 (a
2 + b2 + c2 − ab − ac − bc)
x = a + (b-a)*rand(1,N); x = triangular rnd(a,b,c,1,N);
0.8 0.8
empirical p.d.f.
empirical p.d.f.
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5
1 1
empirical c.d.f.
empirical c.d.f.
0.8 0.8
0.6 0.6
0.4 0.4
0.2 µ=1.0 σ=0.5 0.2 µ=1.0 σ=0.5
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3
x x

Exponential X ∼ E(µ) Laplace X ∼ L(µ, σ 2 )

X∈ R+ , µ∈ R+ X ∈ R, µ ∈ R, σ ∈ R+
1
p.d.f., f(x)
p.d.f., f(x)
1/(e µ)
0 0
0 µ 2µ 3µ 4µ µ-3 σ µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ
0.95 0.97
0.86 0.88
c.d.f., F(x)
c.d.f., F(x)
0.63
0.5
0.12
0.03
0 µ 2µ 3µ 4µ µ - 2σ µ-σ µ µ+ σ µ+2 σ
x x
√
√
1 x 2 |x − µ|

f (x) = exp − f (x) = exp − 2
µ µ  2σ √ σ
1 |x−µ|

x

2 exp 2 x<µ
F (x) = 1 − exp − F (x) = √ σ
µ 1
 1 − exp − 2 |x−µ|
2 σ x≥µ
µX = µ µX = µ
2 = µ2
σX 2 = σ2
σX
x = exp rnd(muX,1,N); x = laplace rnd(muX,sigmaX,1,N);
0.8
1
empirical p.d.f.
empirical p.d.f.
0.8 0.6
0.6 0.4
0.4
0.2
0.2
0 0
0 1 2 3 4 5 -2 -1 0 1 2 3 4
1 1
empirical c.d.f.
empirical c.d.f.
0.8 0.8
0.6 0.6
0.4 0.4
0.2 µ=1.0 0.2 µ=1.0 σ=1.0
0 0
0 1 2 3 4 5 -2 -1 0 1 2 3 4
x x

Normal X ∼ N (µ, σ 2 ) 2 )
Log-Normal ln X ∼ N (µln X , σln X
p.d.f., f(x)
X ∈ R, µ ∈ R, σ ∈ R+ X ∈ R , µln X ∈ R , σln X ∈ R+
+ +
p.d.f., f(x)
0 0
µ+3 σ µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ 0 µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
0.98 0.96
0.84 0.86
c.d.f., F(x)
c.d.f., F(x)
0.59
0.5
0.16 0.11
0.02
µ+3 σ µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ 0 µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
x x
! !
1 (x − µ)2 1 (ln x − µln X )2
f (x) = √ exp − f (x) = q exp − 2
2πσ 2 2σ 2 2
x 2πσln 2σln X
X
  
1 x−µ 1 ln x − µln X 

F (x) = 1 + erf √ F (x) = 1 + erf  q
2 2σ 2 2 2σ 2 ln X
q
µX = µ µX = xm 1 + c2X
2 = σ2 2 = x2 c2 1 + c2

σX σX m X X
µln X = ln xm = ln µX − 21 σln
2
X
2
σln X = ln 1 + (σX /µX ) 2
x = muX + sigmaX*randn(1,N); x = logn rnd(Xm,Cx,1,N);
0.8
1
empirical p.d.f.
empirical p.d.f.
0.6 0.8
0.4 0.6
0.4
0.2
0.2
0 0
-1 0 1 2 3 0 0.5 1 1.5 2 2.5 3
1 1
empirical c.d.f.
empirical c.d.f.
0.8 0.8
0.6 0.6
0.4 0.4
0.2 µ=1.0 σ=0.5 0.2 µ=1.0 σ=0.5
0 0
-1 0 1 2 3 0 0.5 1 1.5 2 2.5 3
x x
Rayleigh X ∼ R(m) Gamma X ∼ Γ(k, θ)

p.d.f., f(x)
X ∈ R+ , m ∈ R+ X ∈ R+ , µ ∈ R+ , σ ∈ R+
p.d.f., f(x)
0 0
0 µ-σ m µ µ+ σ µ+2 σ µ+3 σ µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
0.96 0.96
0.84 0.85
c.d.f., F(x)
c.d.f., F(x)
0.54 0.57
0.16 0.14
0
0 µ-σ m µ µ+ σ µ+2 σ µ+3 σ µ - 2σ µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
x x
!
x 1 x 2 1 x

f (x) = 2 exp − f (x) = k
xk−1 exp −
m 2 m ! Γ(k)θ θ
1 x 2 1 x

F (x) = 1 − exp − F (x) = γ k,
2 m Γ(k) θ
p
µX = m π/2 µX = kθ
2 = m2 (4 − π)/2
σX 2 = kθ 2
σX
x = Rayleigh rnd(modeX,1,N); x = gamma rnd(muX,covX,1,N);
1 1
0.8 0.8
empirical p.d.f.
empirical p.d.f.
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
1 1
empirical c.d.f.
empirical c.d.f.
0.8 0.8
0.6 0.6
0.4 0.4
0.2 µ=1.0 0.2 µ=1.0 σ=0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
x x

Extreme I X ∼ EI [m, s] Extreme II X ∼ EII (m, s, k)

p.d.f., f(x)
X ∈ R, m ∈ R, s ∈ R+ , X > m, m ∈ R, s ∈ R+ , k > 0, k 6= 1, 2
p.d.f., f(x)
0 0
µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ 0 µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
0.96 0.96
0.86 0.9
c.d.f., F(x)
c.d.f., F(x)
0.65
0.57
0.13
0 0.01
µ-2 σ µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ 0 µ-σ µ µ+ σ µ+2 σ µ+3 σ µ+4 σ
x x
−1−k " −k #

1 x−m x−m k x−m x−m

f (x) = exp − + exp − f (x) = exp −
s s s s s s
x−m −k
( h i
x−m

exp −

s x>m
F (x) = exp − exp − F (x) =
s 0 otherwise
µX = m + γs , γ ≈ 0.5772 µX = m + sΓ(1 − 1/k)

2 = (π 2 /6) s2
σX 2 = s2 [Γ(1 − 2/k) − (Γ(1 − 1/k))2 ]
σX
x = extI rnd(mu,cv,1,N); x = extII rnd(m,s,k,1,N);
1 2
0.8
empirical p.d.f.
empirical p.d.f.
1.5
0.6
1
0.4
0.5
0.2
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
1 1
empirical c.d.f.
empirical c.d.f.
0.8 0.8
0.6 0.6
0.4 0.4
0.2 µ=1.0 σ=0.5 0.2 µ=1.0 σ=0.5
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
x x

4 Sums and Differences of Independent Normal Random Variables
Consider two normally distributed random variables, X ∼ N (µX , σX2 ) and Y ∼ N (µ , σ 2 ). Any
Y Y
weighted sum of normal random variables is also normally distributed.
Z = aX − bY

Z ∼ N aµX − bµY , (aσX )2 + (bσY )2
µZ = aµX − bµY σZ2 = (aσX )2 + (bσY )2
5 Products and Quotients of Independent LogNormal Random Variables

2 ) and ln Y ∼ N (µ
Consider two log-normally distributed random variables, ln X ∼ N (µln X , σln 2
X ln Y , σln Y ).
Any product or quotient of lognormal random variables is also lognormally distributed.
Z = X/Y

2 2
ln Z ∼ N µln X − µln Y , σln X + σln Y
2 2 2
µln Z = µln X − µln Y σln Z = σln X + σln Y c2Z = c2X + c2Y + c2X c2Y
6 Examples
1. The strength, S, of a particular grade of steel is log-normally distributed with median 36 ksi
and c.o.v. of 0.15. What is the probability that the strength of a particular sample is greater
than 40 ksi?
ln 40 − ln 36 3.69 − 3.58

P [S > 40] = 1 − P [S ≤ 40] = 1 − Φ =1−Φ
0.15 0.15
= 1 − Φ(0.702) = 1 − 0.759 = 0.241
0.08
smode = 35.20 ksi
0.07 smedian = 36.00 ksi
smean = 36.40 ksi
0.06
0.05
p.d.f.
0.04
0.03
0.02
0.01 P[S>40]
0
20 25 30 35 40 45 50 55 60
strength of steel, s, ksi

2. Highway truck weights in Michigan, W , are assumed to be normally distributed with mean
100 k and standard deviation 40 k. The load capacity of bridges in Michigan, R, are also
assumed to be normally distributed with mean 200 k and standard devation 30 k. What is
the probability of a truck exceeding a bridge load rating?
E = W − R. If E > 0 the truck weight eceededs the bridge capacity.
µW − µR = 100 − 200 = −100 k.
µE = √
σE = 402 + 302 = 50 k.
0 − (−100)

P [E > 0] = 1 − P [E ≤ 0] = 1 − Φ = 1 − Φ(2) = 1 − 0.977 = 0.023
50
R
W
E=W−R 40 30
50 000000000000000
111111111111111
000000000000000
111111111111111
−100 0 100 200 k
3. Windows in the Cape Hattaras Lighthouse can withstand wind pressures of R. R is log-
normal with median of 40 psf and coefficient of variation of 0.25. The peak wind pressure
during a hurricane P in psf is given by the equation P = 1.165 × 10−3 CV 2 where C is a
log-normal coefficient with median of 1.8 and coefficient of variation of 0.20 and V is the
wind speed with median 100 fps and coefficient of variation of 0.30. What is the probability
of the wind pressure exceeding the strength of the window?
The peak wind pressure is also log-normal.
ln P = ln(1.165 × 10−3 ) + ln C + 2 ln V
µln P = ln(1.165 × 10−3 ) + µln C + 2µln V
µln P = ln(1.165 × 10−3 ) + ln(1.8) + 2 ln(100) = 3.0431
2
σln P = ln(1 + 0.202 ) + 2 ln(1 + 0.302 ) = 0.2116 . . . σln P = 0.4600
The wind pressure exceeds the resistance if P/R > 1 (that is, if ln P − ln R > 0)
ln E = ln P − ln R
µln E = µln P − µln R = 3.0431 − ln(40) = −0.646
2 2
σln E = 0.2116 + ln(1 + 0.25 ) = 0.27220 . . . σln E = 0.5217
The probability of the wind load load exceeding the resistance of the glass is,
0 + 0.646

P [E > 1] = 1 − P [E ≤ 1] = 1 − P [ln E ≤ 0] = 1 − Φ = 1 − Φ(1.2383) = 0.11
0.5217
4. When a strong earthquake (i.e., M > 6) occurs, the intensity of the ground shaking (e.g.,
peak ground accelration) at a nearby building site is uncertain. Assume that the peak ground
acceleration (PGA) from a strong earthquake is log-normally distributed with median of 0.2 g
and a coefficient of variation of 0.25. Assume that the building will sustain no damage for
ground motion shaking up to 0.3 g. What is the probability the building will be damaged
from an earthquake of M > 6?
ln(0.3) − ln(0.2)

P [D|M > 6] = P [P GA > 0.3] = 1−P [P GA ≤ 0.3] = 1−Φ = 1−0.947 = 0.053.
0.25

There have been two earthquakes with M > 6 in the last 50 years. What is the probability
of no damage from earthquakes with M > 6 in the next 20 years?
P [D0 |M > 6] = 1 − 0.053 = 0.947
From the theorem of total probability,
P [D0 ] in 20 yr = P [D0 |0 EQ M > 6 in 20yr] · P [0 EQ M > 6 in 20yr] +

P [D0 |1 EQ M > 6 in 20yr] · P [1 EQ M > 6 in 20yr] +
P [D0 |2 EQ M > 6 in 20yr] · P [2 EQ M > 6 in 20yr] +
P [D0 |3 EQ M > 6 in 20yr] · P [3 EQ M > 6 in 20yr] + · · ·
where P [D0 |n EQ M > 6] = (P [D0 |1 EQ M > 6])n (assuming damage from an earthquake
does not weaken the building . . . ) So,
∞
(20/25)n
P [D0 ] in 20 yr
X
= (0.947)n exp(−20/25)
n=0
n!
" #
0.8 0.82 0.83
= exp(−0.8) 1 + 0.947 + (0.947)2 + (0.947)3 + ···
1! 2! 3!
= exp(−0.8) · exp(0.947 · 0.8) = 0.958
So the probability of damage from earthquakes in the next 20 years (given the assumptions
in this example) is close to 4%. Is 4% in 20 years an acceptable level of risk to you for the
possibility of earthquake damage to your home?
7 Empirical PDFs, CDFs, and exceedence rates (nonparametric statistics)

The PDF and CDF of a sample of random data can be computed directly from the sample, without
assuming any particular probability distribution . . . (such as a normal, exponential, or other kind
of distribution).
A random sample of N data points can be sorted into increasing numerical order, so that
x1 ≤ x2 ≤ · · · ≤ xi−1 ≤ xi ≤ xi+1 ≤ · · · ≤ xN −1 ≤ xN .
In the ordered sample there are i data points less than or equal to xi . So, if the sample is represen-
tative of the population, and the sample is “big enough” the probability that a random X is less
than or equal to the ith ordered value is i/N . In other words, P [X ≤ xi ] = i/N . Unless we know
that no value of X can exceed xN , we must accept some probability that X > xN . So, P [X ≤ xN ]
should be less than 1. In such cases, the unbiased estimate1 E[FX (xi )] for P [X ≤ xi ] is i/(N + 1)
The empirical CDF computed from a ordered sample of N values is

i
F̂X (xi ) =
N +1
The empirical PDF is basically a histogram of the data. The following Matlab lines plot empirical
CDFs and PDFs from a vector of random data, x.
1
E.J. Gumbel, Statistics of extremes, Columbia Univ Press, 1958
Lasse Makkonen, “Problems in the extreme value analysis,” Structural Safety 2008:30:405-419

1 N = length ( x ); % number o f v a l u e s i n t h e sample

2 nBins = f l o o r ( N /50); % number o f b i n s i n t h e h i s t o g r a m
3 [ fx , xx ] = h i s t (x , nBins ); % compute t h e h i s t o g r a m
4 fx = fx / N * nBins /(max( x ) -min( x ))); % s c a l e t h e h i s t o g r a m t o a PDF
5 F_x = ([1: N ])/( N +1); % e m p i r i c a l CDF
6 subplot (211); bar( xx , fx ); % p l o t e m p i r i c a l PDF
7 subplot (212); s t a i r s ( sort ( x ) , F_x ); % p l o t e m p i r i c a l CDF
8 p r o b a b i l i t y _ o f _ f a i l u r e = sum(x >0) / N % probability that X > 0
The number of values in the sample greater than xi is (N − i). If the sample is representative, the
probability of a value exceeding xi is Prob[X > xi ] = 1 − FX (xi ) ≈ 1 − i/N . If the N samples were
collected over a period of time T , the average exceedence rate (number of events greater than xi
per unit time) is ν(xi ) = N (1 − FX (xi ))/T ≈ N (1 − i/N )/T = (N − i)/T .
8 Random variable generation using the Inverse CDF method

A sample of a random variable having virtually any type of CDF, P [X ≤ x] = P = FX (x) can
be generated from a sample of a uniformly distributed random variable, U , (0 < U < 1), as
long as the inverse CDF, x = FX−1 (P ) can be computed. There are many numerical methods for
generating a sample of uniformly distributed random numbers. It is important to be aware that
samples from some methods are “more random” than samples from others. The Matlab command
u = rand(1,N) computes a (row) vector sample of N uniformly distributed random numbers with
0 < u < 1.
If X is a continuous rv with CDF FX (x) and U has a uniform distribution on (0, 1), then the
random variable FX−1 (U ) has the distribution FX . Thus, in order to generate a sample of data
distributed according to the CDF FX , it suffices to generate a sample, u, of the rv U ∼ U[0, 1] and
then make the transformation x = FX−1 (u).
For example, if X is exponentially distributed, the CDF of X is given by

FX (x) = 1 − e−x/µ ,
so
FX−1 (u) = −µ ln(1 − FX (x)).
Therefore if u is a value from a uniformly distributed rv in [0, 1], then
x = −µ ln(u)
is a value from an exponentially distributed random variable. (If U is uniformly distributed in [0,1]
then so is 1 − U .)
As another example, if X is log-normally distributed, the CDF of X is

ln x − ln xm

FX (x) = Φ .
σln X
If u is a sample from a standard uniform distribution, then
h i
x = exp ln xm + Φ−1 (u)σln X
is a sample from a lognormal distribution.
Note that since expressions for Φ(z) and Φ−1 (P ) do not exist, the generation of normally distributed
random variables requires other numerical methods. x = muX + sigX*randn(1,N) computes a
(row) vector sample of N normally distributed random numbers.

cdf cdf
1
1
F(x) F(x)
X X
u
u
u = F (x) u = F (x)
X X
f (u)
f (u)
−1
pdf
−1
pdf
x = F (u) x = F (u)
U
U
X X
0
0
x=a x=b x x=a x=b x
1
1
pdf pdf
dFX dFX
f (x) = f (x) =
X dx X dx
x=a x=b x x=a x=b x
Figure 1. Examples of the generation of uniform random variables from the inverse CDF method.
u
cdf cdf
u = FX (x) 1 u = FX(x)
1
F(x) F(x)
X
X
−1 −1
f (u)
f (u)
x = F (u) x = FX (u)
pdf
pdf
U
X
0
0
x x
1
pdf pdf
dFX dFX
f (x) = f (x) =
X dx X dx
x x
Figure 2. Examples of the generation of random variables from the inverse CDF method. The density of
−1
the horizontal arrows u is uniform, whereas the density of the vertical arrows, x = FX (u), is proportional
0
to FX (x), that is, proportional to fX (x).

9 Correlated Random Variables

If the cumulative distribution function FX (x) is invertible, the random variable X can be related
to a standard normal random variable Y through ΦY (y) = FX (x), from which a sample of values
of Y can be calculated from a sample of values of X as
[y1 , y2 , ..., ym ] = Φ−1 (FX ([x1 , x2 , ..., xm ]))
and a sample of values of X can be calculated from a sample of values of Y as
[x1 , x2 , ..., xm ] = F −1 (ΦY ([y1 , y2 , ..., ym ]))
A sample of m values of a set of n correlated standard normal random variables {Y1 , Y2 , ..., Yn }
with a given correlation matrix Rn×n can be computed from a sample of m values of a set of n
uncorrelated standard normal variables {Z1 , Z2 , ..., Zn } from the eigenvalues and eigenvectors of R,
R = V ΛV T .
   
y11 y12 y13 ··· y1m z11 z12 z13 ··· z1m

y21 y22 y23 ··· y2m  
z21 z22 z23 ··· z2m 
 = V Λ1/2 
   
 .. .. .. .. .. .. .. .. 
. . ··· . . . . ··· . .
   
   
yn1 yn2 yn3 · · · ynm zn1 zn2 zn3 · · · znm
where yij is sample j from random variable Yi and zij is sample j from random variable Zi .
To compute this in Matlab,
[eVec,eVal] = eig(R);
Z = randn(n,m);
Y = eVec * sqrt(eVal) * Z;
X = inv_cdf(normcdf(Y),a,b);
where cdf_inv() is a particular inverse cumulative distribution function such as

logn_inv(normcdf(Y),medianX,covX); or Rayleigh_inv(normcdf(Y),modeX); .
10 Functions of Random Variables and Monte Carlo Simulation

The probability distributions of virtually any function of random variables can be computed using
the powerful method of Monte Carlo Simulation (MCS). MCS involves computing values of functions
with large samples of random variables.
For example, consider a function of three random variables, X1 , X2 , and X3 , where X1 is normally
distributed with mean of 6 and standard deviation of 2, X2 is log-normally distributed with median
of 2 and coefficient of variation of 0.3, and X3 is Rayleigh distributed with mode of 1. The function
p
Y = sin(X1 ) + X2 − exp(−X3 ) − 2
is a function of these three random variables and is therefore also random. The distribution function
and statistics of Y may be difficult to derive analytically, especially if the function Y = g(X) is
complicated. This is where MCS is powerful. Given samples of N values of X1 , X2 and X3 , a
sample of N values of Y can also be computed. The statistics of Y (mean, variance, PDF, and
CDF) can be estimated by computing the average value, sample variance, histogram, and emperical
CDF of the sample of Y . The probability P [Y > 0] can be estimated by counting the number of
positive values in the sample and dividing by N . The Matlab command P_Y_gt_0 = sum(y>0)/N
may be used to estimate this probability.

11 Monte Carlo Simulation in Matlab
1 % MCS intro .m
2 % Monte Car lo S i m u l a t i o n . . . an i n t r o d u c t o r y example
3 %
4 % Y = g (X1 , X2 , X3) = s i n (X1) + s q r t (X2) − exp(−X3) − 2 ;
5 %
6 % H. P . Gavin , Dept . C i v i l and Environmental E n g i n e e r i n g , Duke Univ , Jan . 2012
7
8 % X1 X2 X3
9 % normal lognormal Rayleigh
10 mu1 = 6; med2 = 2; mod3 = 1;
11 sd1 = 2; cv2 = 0.3;
12
13 N = 1000; % number o f random v a l u e s i n t h e sample
14
15 % ( 1 ) g e n e r a t e a l a r g e sample f o r each random v a r i a b l e i n t h e problem . . .
16
17 X1 = mu1 + sd1 *randn(1 , N );
18 X2 = logn_rnd ( med2 , cv2 ,1 , N );
19 X3 = Rayleigh_rnd ( mod3 ,1 , N );
20
21 % ( 2 ) e v a l u a t e t h e f u n c t i o n f o r each random v a r i a b l e t o compute a new sample
22
23 Y = s i n ( X1 ) + sqrt ( X2 ) - exp( - X3 ) - 2;
24
25 % s u p p o s e ” p r o b a b i l i t y o f f a i l u r e ” i s Prob [ g (X1 , X2 , X3) > 0 ] ...
26
27 P r o b a b i l i t y _ o f _ f a i l u r e = sum(Y >0) / N
28
29 % ( 3 ) p l o t h i s t o g r a m s o f t h e random v a r i a b l e s
30
31 sort_X1 = sort ( X1 );
34 CDF = [1: N ] / (1+ N ); % e m p i r i c a l CDF o f a l l quantities

0.6
0.6
0.5
0.15
0.4
P.D.F.
0.4
0.1 0.3
0.2
0.2
0.05
0.1
0 0 0
0 2 4 6 8 10 1 2 3 4 5 0.5 1 1.5 2 2.5 3 3.5
0.8 0.8 0.8
0.6 0.6 0.6

C.D.F.
0.4 0.4 0.4
0.2 0.2 0.2
0 2 4 6 8 10 1 2 3 4 5 0.5 1 1.5 2 2.5 3 3.5

X : normal X : log-normal X : Rayleigh
1 2 3
0.8
0.6
P.D.F.
0.4
0.2
0
-3 -2 -1 0 1
0.8
0.6
C.D.F.
Y>0
0.4
0.2
0
-2.5 -2 -1.5 -1 -0.5 0 0.5
Y = g(X ,X ,X )
1 2 3
Figure 3. Analytical and empirical PDF’s and CDF’s for X1 , X2 , and X3 , and the Empirical PDF and
CDF for Y = g(X1 , X2 , X3 )
1 nBins = f l o o r ( N /50);
2 f i g u r e (1)
3 clf
4 subplot (231)
5 [ fx , xx ] = h i s t ( X1 , nBins , nBins /(max( X1 ) -min( X1 )) ); % h i s t o g r a m o f X1
6 hold on
7 bar( xx , fx , ’ FaceColor ’ ,[1 1 1])
8 plot ( sort_X1 , normpdf ( sort_X1 , mu1 , sd1 ) , ’ LineWidth ’ ,2);
9 hold off
10 axis ( ’ tight ’)
11 ylabel ( ’P . D . F . ’)
12 subplot (234)
13 hold on
14 s t a i r s ( sort_X1 , CDF , ’ -b ’ , ’ LineWidth ’ ,2)
15 plot ( sort_X1 , normcdf ( sort_X1 , mu1 , sd1 ) , ’ -r ’ )
16 hold off
18 ylabel ( ’C . D . F . ’)
19 xlabel ( ’ X_1 : normal ’)
20 subplot (232)
22 hold on
23 bar( xx , fx , ’ FaceColor ’ ,[1 1 1]);
24 plot ( sort_X2 , logn_pdf ( sort_X2 , med2 , cv2 ) , ’ LineWidth ’ ,2);
25 hold off
27 subplot (235)
28 hold on
30 plot ( sort_X2 , logn_cdf ( sort_X2 ,[ med2 , cv2 ]) , ’ -r ’ )
31 hold off
33 xlabel ( ’ X_2 : log - normal ’)
34 subplot (233)
36 hold on
38 plot ( sort_X3 , Rayleigh_pdf ( sort_X3 , mod3 ) , ’ LineWidth ’ ,2 );
39 hold off
41 subplot (236)
42 hold on
44 plot ( sort_X3 , Rayleigh_cdf ( sort_X3 , mod3 ) , ’ -r ’ )
45 hold off
47 xlabel ( ’ X_3 : Rayleigh ’)
48 nBins = f l o o r ( N /20);
49 f i g u r e (2)
50 clf
51 subplot (211)
52 [ fx , xx ] = h i s t ( Y , nBins , nBins /(max( Y ) -min( Y )) ); % histogram of Y
53 hold on
55 plot ([0 0] ,[0 0.5] , ’ -k ’ , ’ LineWidth ’ ,3)
56 hold off
58 ylabel ( ’P . D . F . ’)
59 subplot (212)
60 hold on
61 s t a i r s ( sort ( Y ) , CDF )
62 plot ( [0 0] ,[0 1] , ’ -k ’ , ’ LineWidth ’ ,3);
63 hold off
65 text (0.5 ,0.5 , ’Y >0 ’)
66 ylabel ( ’C . D . F . ’)
67 xlabel ( ’Y = g ( X_1 , X_2 , X_3 ) ’ );

Probability Distributions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Probability Distributions

Uploaded by

Copyright:

Available Formats

Probability Distributions

CEE 201L. Uncertainty, Design, and Optimization

1. The PDF fX (x) is positive-valued; fX (x) ≥ 0 for all values of x ∈ X .

A few important characteristics of CDF’s of X are:

1. CDF’s, FX (x), are monotonic non-decreasing functions of x.

2 Descriptors of random variables

The median value, xm , is is the value of x such that

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

3 Some common distributions

3.1 The Normal distribution

If X is normally distributed, it has the PDF

and more specifically, aX − b ∼ N aµX − b, a2 σX

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

3.2 The Log-Normal distribution

If ln X is normally distributed (i.e., ln X ∼ N (µln X , σln X )) then X is called a log-normal random

µY = µln X , σY2 = σln

If ln X is normally distributed (X is log-normal) then (for cX < 0.3)

where µln Z = ln a + nµln X − mµln Y = ln a + n ln xm − m ln ym

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

Uniform X ∼ U[a, b] Triangular X ∼ T (a, b, c)

a µ-σ µ µ+ σ b a µ-σ c µ µ+ σ µ+2 σ b

x = a + (b-a)*rand(1,N); x = triangular rnd(a,b,c,1,N);

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

Exponential X ∼ E(µ) Laplace X ∼ L(µ, σ 2 )

x = exp rnd(muX,1,N); x = laplace rnd(muX,sigmaX,1,N);

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

x = muX + sigmaX*randn(1,N); x = logn rnd(Xm,Cx,1,N);

Rayleigh X ∼ R(m) Gamma X ∼ Γ(k, θ)

x = Rayleigh rnd(modeX,1,N); x = gamma rnd(muX,covX,1,N);

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

Extreme I X ∼ EI [m, s] Extreme II X ∼ EII (m, s, k)

−1−k " −k #

µX = m + γs , γ ≈ 0.5772 µX = m + sΓ(1 − 1/k)

x = extI rnd(mu,cv,1,N); x = extII rnd(m,s,k,1,N);

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

4 Sums and Differences of Independent Normal Random Variables

µZ = aµX − bµY σZ2 = (aσX )2 + (bσY )2

5 Products and Quotients of Independent LogNormal Random Variables

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

P [D0 |M > 6] = 1 − 0.053 = 0.947

From the theorem of total probability,

P [D0 ] in 20 yr = P [D0 |0 EQ M > 6 in 20yr] · P [0 EQ M > 6 in 20yr] +

7 Empirical PDFs, CDFs, and exceedence rates (nonparametric statistics)

The empirical CDF computed from a ordered sample of N values is

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

1 N = length ( x ); % number o f v a l u e s i n t h e sample

8 Random variable generation using the Inverse CDF method

For example, if X is exponentially distributed, the CDF of X is given by

As another example, if X is log-normally distributed, the CDF of X is

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

x=a x=b x x=a x=b x

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

9 Correlated Random Variables

To compute this in Matlab,

where cdf_inv() is a particular inverse cumulative distribution function such as

10 Functions of Random Variables and Monte Carlo Simulation

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

11 Monte Carlo Simulation in Matlab

CC BY-NC-ND March 25, 2021 PSH, HPG, JTS

0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

−1−k " −k #