On modeling correlated random variables in risk assessment

Area of Science:

• Quantitative risk assessment
• Computational statistics

Background:

• Monte Carlo methods are crucial for risk assessment.
• Incorporating input correlations is vital for accurate simulations.
• The Iman and Conover method is a common approach for generating correlated variables.

Purpose of the Study:

• To introduce alternative methods for generating correlated random variables in Monte Carlo simulations.
• To evaluate the impact of different correlation generation techniques on risk assessment outcomes.

Main Methods:

• Utilized copula functions to derive correlated random variables.
• Compared copula-based methods with the Iman and Conover method.
• Analyzed the influence of higher-order bivariate moments on simulation results.

Main Results:

• Copula-based methods provide an alternative to the Iman and Conover technique.
• The choice of correlation generation method can significantly alter simulation outputs.
• Differences in higher-order moments between methods lead to varied results.

Conclusions:

• Copula methods offer a flexible approach for modeling input dependencies in risk assessment.
• Careful selection of correlation generation techniques is essential for reliable Monte Carlo risk analysis.