Monte Carlo (MC) simulation is the name given to a method of analysis that runs iterations to produce a statistical result. The process is opposed to deriving a function that would give the required result (often an almost impossible task in anything but the simplest of plans).
Latin Hypercube (LH) simulation is a stratified sampling method that gives outcomes nearer the input distribution theoretical values with less iteration. The two simulations do it differently, but many risk analysts call all simulations Monte Carlo.
It is not a big deal to those who are not profoundly intertwined with the risk analysis process. Some risk professionals will call your attention and make the correction if you use Latin Hypercube in simulation but keeps calling it Monte Carlo. The good thing is, it will not drastically change the result. We want to differentiate one process from another.
A risk manager should be precise with the terminology if he can help it. The efficacy of Monte Carlo (MC) or Latin Hypercube (LH) has nothing to do with the type of model (e.g., cost estimate or schedule). It is practically just a question of how much time it takes to reach stability and develop the results.
Generally, LH stabilizes results quicker than MC. However, a Monte Carlo simulation of a schedule with few activities could stabilize quicker than a Latin Hypercube. MC takes the three points to estimate each activity (represented by the triangle). The tool samples a random segment in each iteration. Randomness makes it less likely but possible that the same sample segment is selected more than once.
MC versus LH iteration
Each iteration is a “what-if” scenario of the project. One thousand iterations are like looking at 1000 project scenarios. This means that a thousand iterations look at the same project one thousand times, with a different situation each time. Five thousand iterations are like looking at the same project using five thousand times what-ifs. The value of each three points range contributes to the overall probability distribution.
Latin Hypercube does the same, but without repeating previously sampled values. Each sample is unique. To conclude, the black box sampling method provides a better representation of the risk model. It is also relatively faster even though we are really talking only time units in seconds.
About the Author
Rufran C. Frago is the Founder of PM Solution Pro, a Calgary consulting, product, and training services firm focusing on project and business management solutions. He is passionate providing advice, mentorship, education and training through consultation, collaboration, and what he uniquely calls, student-led training.
BOOKS AUTHORED BY RUFRAN FRAGO
- Risk-based Management in the World of Threats and Opportunities: A Project Controls Perspective.ISBN 978-0-9947608-0-7.Canada
- Plan to Schedule, Schedule to Plan.ISBN 978-0-9947608-2-1.Canada
- How to Create a Good Quality P50 Risk-based Baseline Schedule.ISBN 978-0-9947608-1-4.Canada
- Schedule Quantitative Risk Analysis (Traditional Method).ISBN 978-0-9947608-3-8.Canada
- RISK, What are you? The Risk Management Poem: Children's Book for all Professionals.ISBN 978-0-9947608-4-5 (Canada)