SIMULATION MODELS IN

INDUSTRIAL ENGINEERING
Lesson 5
Random-Variate Generation

Tran Van Ly
Industrial Engineering & Management
International University
Email: tvly@hcmiu.edu.vn 1
Room: A2-504
 Purpose

• Develop understanding of generating

samples from a specified distribution as
input to a simulation model.

• Illustrate some widely-used techniques for

generating random variates.
– Inverse-transform technique
– Acceptance-rejection technique
– Special properties
2
 1. Inverse-transform Technique
1.1. Continuous distribution
• The concept: Suppose we
wish to generate a Random
Variate X that is continuous,
has distribution function cdf F:
0 < F(X) <1. Denote F-1 the
inverse of function F.
• The algorithm is F(x)
1
– For cdf function: R = F(X) R = F(x)
1. Generate R from uniform U(0,1) R

2. Return X:

X = F-1(R) X
3
Fig. 8.1 Inverse Transform Method
 a. Exponential Distribution
• Exponential Distribution:
– Exponential

 xe  x , x  0 1  e  x , x  0
f  x    F  x    f (t )dt 
 0, x 0   0, x 0

– Algorithm
• Generate R ~U(0,1)
• Solve 1-e-λx = R
1
 X   ln(1  R )


• If we want to generate n values of

X:: X1, X2,…, Xn
- Generate uniform RN: R1, R2,…, Rn
- Return: 1
X i   ln(1  Ri ) ; i= 1, 2, …, n
 Figure: Inverse-transform technique for exp( = 1)

4
 b. Uniform Distribution U(a,b)

– Uniform
 1
distribution 0; x  a
 x  a
 , a  xb F  x  
f  x   b  a , a  xb
 b  a
0, elsewhere  1, x b

– Algorithm
• Generate R~U(0,1)
• Return X = a + (b-a)R

5
 c. Weibull Distributions

– Weibull distribution
   1 ( x / a ) 
 x e , x0
f ( x )   
 0 , otherwise
, β are scale and shape parameters
x
( ) 
F  x  1  e 
,x 0

– Algorithm
– Generate R~U(0,1)
– Return 1

X   [ ln(1  R)] 

6
1.2 Empirical Continuous Distributions
• Unable to find theoretical distribution → use empirical distribution
• Procedure:
– Collect data Xi; i= 1, 2, …, n
– Arrange data from smallest to largest
– Assign the probability 1/n to each interval [xi-1, xi]
– Compute the slope of ith line segment
x(i )  x(i 1) x(i )  x(i 1) i/n
ai  
i / n  (i  1) / n 1/ n R

– Generate R
– Return (i  1) 
(i-1)/n


X  Fˆ 1 ( R)  x(i 1)  ai  R   xi-1 x xi
 n 
X  x( i 1)
ai 
R  (i  1) / n 7
1.2 Empirical Continuous Distributions (cont.)
• Five observations of fire-crew response times (in mins.):
– 2.76 1.83 0.80 1.45 1.24
– Arranging x0 = 0; x1 = 0.8; x2 = 1.24; x3 = 1.45; x4 = 1.83; x5 = 2.76
– n = 5  probability for each interval = 1/n = 0.2

Cumalative
Interval Probability Probability,
i (Hours) 1/n i/n Slope, a i
1 0.0 ≤ x ≤ 0.80 0.2 0.20 4.00
2 0.8 ≤ x ≤ 1.24 0.2 0.40 2.20
3 1.24 ≤ x ≤ 1.45 0.2 0.60 1.05
4 1.45 ≤ x ≤ 1.83 0.2 0.80 1.90
5 1.83 ≤ x ≤ 2.76 0.2 1.00 4.65

8
1.2 Empirical Continuous Distributions (cont.)

Consider R 1 = 0.71:

(i-1)/n = 0.6 < R 1 < i/n = 0.8

X 1 = x (4-1) + a 4(R 1 – (4-1)/n)

= 1.45 + 1.90(0.71-0.60)
= 1.66

9
 1. 3 Discrete Distribution

• Procedure P(x)
– With given pmf P(xi) = P(X=xi),
compute cdf
F ( x)  P( X  x)   P( xi )
xi  x

(Assume that X can take only the value x1<x2<… <xn) X1 X2 Xi-1 Xi
– Generate R~U(0,1) 1
– Determine the smallest integer I:
R≤ F(xI) R
Return X = xI
F2

X1 X2 Xi-1 Xi
10
 1. 3 Discrete Distribution
1
p.m. f : p  x   ; x  1..k
k
c.d . f :
0 x 1
1
 k 1 x  2

 2 2 x3
 k
F  x  
...
 k 1
 k -1  x  k
 k Generate R~U(0,1)

1 kx If R satisfy:
i 1 i
F  i  1   R   F  i
k k
Then return X = i
11
 Bernoulli and Binomial Distribution
Bernoulli Distribution
p.m. f : P X  0  p ; P X  1  1  p

• Generate R~U(0,1)
• If R ≤ p return X = 1, otherwise X = 0

Binomial Distribution B(n,p)

• Generate Y1, Y2, … , Yn as Bernoulli (p) random variates.

• Return X = Y1 + Y2 + … + Yn

12
 2. Acceptance-Rejection technique
• Useful particularly when inverse cdf does not exist in closed form, a.k.a.
thinning
• Illustration: To generate random variates, X ~ U(1/4, 1)

Generate R
Procedures:
no
Step 1. Generate R ~ U[0,1]
Condition
Step 2a. If R >= ¼, accept X=R.
yes
Step 2b. If R < ¼, reject R, return
to Step 1 Output R’
• R does not have the desired distribution, but R conditioned (R’) on the
event {R  ¼} does.
• Efficiency: Depends heavily on the ability to minimize the number of
rejections.
13
 Poisson Distribution

Based on the relationship between Poisson(λ) and Expo(1/λ)

•Set n = 0 and P =1
•Generate random number Ri+1 and replace P by

P x Ri+1 .
•If P < e-λ then accept X = i. Otherwise, reject current i, increase i by
1 and return previous step.

14
 3. Special Properties
• Based on features of particular family of
probability distributions
• For example:
– Direct Transformation for normal and lognormal
distributions
– Convolution
– Beta distribution (from gamma distribution)

15
 Direct Transformation [Special Properties]
• Approach for normal(0,1):
– Consider two standard normal random variables, Z1 and Z2, plotted as a
point in the plane:

In polar coordinates:
Z1 = B cos 
Z2 = B sin 

1–/ 2
B2 = Z21 + Z22 ~ chi-square distribution with 2 degrees of freedom =
B  (2 ln R)
Exp( = 2). Hence,
– The radius B and angle  are mutually independent.
Z1  ( 2 ln R )1/ 2 cos(2R2 )
Z 2  ( 2 ln R )1/ 2 sin( 2R2 )
16
 Direct Transformation [Special Properties]
• Approach for normal(,):
– Generate Zi ~ N(0,1)
Xi =  +  Zi

• Approach for lognormal(,):

– Generate X ~ N(,)
Yi = eXi

17
 Summary
• Principles of random-variate generate via
– Inverse-transform technique
– Acceptance-rejection technique
– Special properties
• Important for generating continuous and
discrete distributions

18