Level:
Intermediate
TCM Section(s)
7.6. Risk Management
3.3. Investment Decision Making
Venue: 2024 AACE International Conference & Expo
Abstract: Practical quantitative risk analysis (QRA) methods using Monte Carlo simulation (MCS) for project contingency determination often rely on 3-point distributions. In a typical project QRA workshop, 3-points (e.g., low, most likely, high) have proven to be the extent of the cost and schedule risk impact range input that most analysts can expect to obtain from project teams. In AACE RP 66R-11, there are only two practical probability distribution functions (PDFs) defined by 3-points: Triangular and PERT. Neither is an ideal representation of naturally occurring cost growth and schedule slip data distributions. Industry research suggests that the Lognormal distribution is a more realistic fit for large projects based on actual project cost and schedule variability. Risk analysts, attempting to be realistic, often resort to PDF variation (e.g., trigen) and/or attempt to push teams toward the analyst’s opinion of acceptable 3-point input. This manipulation can raise stakeholder suspicions about the reliability of MCS, and rightfully so.
What is needed is a lognormally shaped PDF that can be defined with 3-point input. Recently, researchers at the University of Texas developed a distribution system called the Johnson quantile-parameterized distribution (J-QPD) that is able to match the shapes of common distributions, including lognormal with 3-point input. This paper will review the typical QRA PDF use context, describe the J-QPD distribution and its use criteria, and provide practical examples of its application in an MCS-based QRA method. The authors will propose an RP 66R-11 update and encourage MCS software vendors to include the J-QPD as a distribution choice in their products.
TCM Section(s)
7.6. Risk Management
3.3. Investment Decision Making
Venue: 2024 AACE International Conference & Expo
Abstract: Practical quantitative risk analysis (QRA) methods using Monte Carlo simulation (MCS) for project contingency determination often rely on 3-point distributions. In a typical project QRA workshop, 3-points (e.g., low, most likely, high) have proven to be the extent of the cost and schedule risk impact range input that most analysts can expect to obtain from project teams. In AACE RP 66R-11, there are only two practical probability distribution functions (PDFs) defined by 3-points: Triangular and PERT. Neither is an ideal representation of naturally occurring cost growth and schedule slip data distributions. Industry research suggests that the Lognormal distribution is a more realistic fit for large projects based on actual project cost and schedule variability. Risk analysts, attempting to be realistic, often resort to PDF variation (e.g., trigen) and/or attempt to push teams toward the analyst’s opinion of acceptable 3-point input. This manipulation can raise stakeholder suspicions about the reliability of MCS, and rightfully so.
What is needed is a lognormally shaped PDF that can be defined with 3-point input. Recently, researchers at the University of Texas developed a distribution system called the Johnson quantile-parameterized distribution (J-QPD) that is able to match the shapes of common distributions, including lognormal with 3-point input. This paper will review the typical QRA PDF use context, describe the J-QPD distribution and its use criteria, and provide practical examples of its application in an MCS-based QRA method. The authors will propose an RP 66R-11 update and encourage MCS software vendors to include the J-QPD as a distribution choice in their products.