Common Random Distributions in Simulation

Overview of Random Distributions in Simulation

  • Classification of Distributions:

    • Discrete Event Simulation (DES) distributions are categorized into two primary types: discrete distributions and continuous distributions.
  • Discrete Distributions Definition:

    • Generate integer values (e.g., 55, 1010, 2020) or countable values.
    • Applied when modeling discrete counts, such as the number of individuals arriving at a station or the number of events occurring at a specific location.
  • Continuous Distributions Definition:

    • Allow non-integer, fractional, or continuous real-valued outputs.
    • Applied when modeling continuous variables such as dollar amounts spent or time durations where fractional values are valid.

Discrete Distributions

  • Bernoulli Distribution:

    • Represents the simplest discrete probability distribution with exactly two possible outcomes (e.g., heads or tails, success or failure).
    • Conceptual Example: Flipping a single coin once to observe whether the outcome is a head or a tail.
  • Binomial Distribution:

    • Involves repeating an experiment multiple times across independent trials and counting the total number of times a specific event occurs (successes).
    • Conceptual Examples:
    • Flipping a coin 2020 times and counting how many times heads appears on top.
    • Rolling a die repeatedly and counting the total number of times the value 66 appears on top.
    • Key Identifying Signature: A problem statement clearly indicates a binomial or Bernoulli distribution if it provides both a probability of success (pp) and a number of trials (nn).
  • Poisson Distribution:

    • Summarizes the total number of times an event occurs within a specified frame of time or space.
    • Expressed with a single primary parameter representing the average or mean rate of occurrence.
    • Practical Examples:
    • The total number of typing errors made while writing a piece of paper.
    • The number of traffic accidents occurring along a specific stretch of road.
  • Uniform Discrete Distribution:

    • Characterized by equal probability across all possible discrete outcomes within a given range.
    • Conceptual Example: Rolling a fair six-sided die, where the probability of observing 11, 22, 33, 44, 55, or 66 are all strictly equal to one another.
    • Applied Scenario: Modeling store customer arrivals expected to be between 4040 and 6060 customers, where every integer value within that range (40,41,42,…,6040, 41, 42, \dots, 60) has an equal probability of occurrence.

Continuous Distributions

  • Continuous Uniform Distribution (genuniform):

    • Differentiated from the discrete uniform distribution by allowing fractional and decimal values.
    • Practical Example: Spending at a store estimated between 1010 dollars and 2020 dollars, where every exact continuous value in between has an equal chance of occurring (e.g., 10.5010.50 dollars, 10.2510.25 dollars, or 10.9110.91 dollars).
    • Distinction from Discrete Models: While spending money allows fractional values like 10.5010.50 dollars, physical discrete count events (such as traffic accidents) cannot take fractional values like 10.510.5 accidents.
  • Triangular Distribution:

    • Frequently used in simulation modeling to represent subjective estimations extracted directly from domain experts in the field.
    • Defined by exactly three parameters:
    • Minimum bound (aa)
    • Maximum bound (bb)
    • Most likely value / Mode (cc)
    • Practical Example: Project duration estimated by a construction foreman stating that a project will take somewhere between 1010 days and 4040 days, but will most likely be completed in 3030 days.
  • Normal Distribution:

    • Symmetric bell-shaped continuous distribution widely used in simulation and statistical modeling.
    • Domain ranges across all real numbers from negative infinity (−∞-\infty) to positive infinity (+∞+\infty).
  • Custom Distribution:

    • Defines custom empirical probabilities for specific discrete outcomes or events.
    • Pairs multiple specific values or events directly with their corresponding individual probabilities of occurrence.

Software Tools and Functions

  • Native Excel Functions:

    • RANDBETWEEN: A native Excel function used to generate discrete uniform random integers between specified lower and upper bounds. Does not require external packages or add-ins.
  • YASAI Software Package:

    • Developed by Rutgers University as a freely available simulation add-in package.
    • Obtainable directly from the original website or course Canvas page.
    • Specialized Functions Included in YASAI:
    • genbinomial: Generates binomial random variates.
    • genpoisson: Generates Poisson random variates.
    • genuniform: Generates continuous uniform random variates.
    • Installation Requirements:
    • Windows Installation: Does not require IT administrator permissions and can be installed independently.
    • macOS Installation: Requires following the dedicated installation guide provided on the official YASAI website.