An efficient method for computing single-parameter partial expected value of perfect information.

Area of Science:

• Decision Analysis
• Computational Statistics
• Risk Management

Background:

• Quantifying the value of learning uncertain input in decision models is essential for informed decision-making.
• The partial expected value of perfect information (EVPI) is a key metric for this quantification.
• Current estimation methods, like 2-level nested Monte Carlo, face implementation challenges with complex conditional distributions.

Purpose of the Study:

• To introduce a novel, simplified 1-level method for calculating partial EVPI for a single parameter.
• To overcome the implementation difficulties associated with sampling conditional distributions in existing 2-level methods.
• To demonstrate the statistical and computational advantages of the proposed 1-level approach.

Main Methods:

• Development of a 1-level Monte Carlo procedure for partial EVPI estimation.
• Derivation of the sampling distribution for the new estimator.
• Comparative analysis against the traditional 2-level nested Monte Carlo method.

Main Results:

• The proposed 1-level method effectively calculates partial EVPI without direct sampling of conditional distributions.
• The sampling distribution of the new estimator was rigorously derived.
• Case study results indicate the 1-level method is statistically and computationally superior to the 2-level method.

Conclusions:

• A simpler and more efficient 1-level method for partial EVPI calculation is presented.
• This new method enhances the practical application of EVPI analysis in decision modeling.
• The findings suggest a significant improvement in both accuracy and computational performance for EVPI estimation.