Random Variate Generation
1. Introduction to Random Variate Generation
Random variate generation is a crucial aspect of simulation modeling. It involves transforming
uniformly distributed random numbers into random variables that follow specific probability
distributions. These distributions model real-world processes such as customer arrivals, service
times, and equipment failures.
Key Concepts:
Random Variate: A random variable generated to follow a specific probability
distribution.
Probability Distribution: A function describing how probabilities are distributed over
possible values of a random variable.
Random Number vs. Random Variate: A random number is uniformly distributed,
whereas a random variate follows a specified distribution.
Illustration:
Consider a bank where customer arrivals follow an exponential distribution. We need a
technique to generate random inter-arrival times based on the exponential distribution.
2. Common Probability Distributions in Simulation
Different distributions model different types of random events. Below are some common
distributions used in simulation:
2.1 Discrete Distributions
1. Bernoulli Distribution:
o Models a binary outcome (success/failure).
o Example: Coin toss (Heads = 1, Tails = 0).
2. Binomial Distribution:
o Models the number of successes in a fixed number of independent trials.
o Example: Suppose a factory produces 100 light bulbs per batch, with a 5% defect
rate. The probability of exactly 3 defective bulbs in a batch is given by: P(X=3)=
3. Poisson Distribution:
o Models the number of events occurring in a fixed interval.
o Example: If a call center receives an average of 5 calls per hour, the probability of
receiving exactly 3 calls in an hour is:
2.2 Continuous Distributions
1. Uniform Distribution:
o All values within a range are equally likely.
2. Example: A random number between 10 and 20 can be generated as:
Exponential Distribution:
o Models time between occurrences of Poisson-distributed events.
o Example: If customer arrivals follow an exponential distribution with λ=4\lambda
= 4 (i.e., an average of 4 arrivals per hour), the probability of a wait time greater
than 15 minutes is:
3.
Normal (Gaussian) Distribution:
o Models natural variations around a mean.
o Example: If exam scores are normally distributed with mean 70 and standard
deviation 10, the probability of scoring above 85 is:
3. Methods for Generating Random Variates
There are several methods to transform uniform random numbers into random variates following
a given distribution.
3.1 Inverse Transform Method
3.2 Acceptance-Rejection Method
Used when the inverse transform method is difficult to apply.
This method involves generating candidate values from an easy-to-sample distribution
and then accepting or rejecting them based on a probability criterion, ensuring that the
final sample follows the desired distribution.
3.3 Box-Muller Method
Specifically designed to generate normal (Gaussian) random variables.
Uses two uniform random numbers U1U_1 and U2U_2 to generate two independent
standard normal variates: Z1=−2lnU1cos(2πU2)Z_1 = \sqrt{-2 \ln U_1} \cos(2\pi U_2)
Z2=−2lnU1sin(2πU2)Z_2 = \sqrt{-2 \ln U_1} \sin(2\pi U_2)
Illustration:
A step-by-step diagram of the inverse transform method and Box-Muller method.
4. Performance Considerations in Random Variate
Generation
Since simulations often require large numbers of random variates, efficiency and accuracy are
critical.
4.1 Computational Efficiency
Some methods (e.g., inverse transform) are faster for specific distributions.
The acceptance-rejection method can be slower if many samples are rejected.
4.2 Accuracy and Precision
Poorly designed PRNGs can produce biased results.
Some methods (e.g., Box-Muller) generate more precise values for normal distributions.
4.3 Reproducibility and Seeding
Using a fixed seed in PRNGs ensures reproducibility.
Important for debugging and verification.
