Related Experiment Videos
Measuring decision sensitivity: a combined Monte Carlo-logistic regression approach
J F Merz1, M J Small, P S Fischbeck
1Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213.
Summary
This study introduces a novel method using stochastic simulation and logistic regression to quantify input variable importance in complex decision models. This approach simplifies sensitivity analysis for better decision-making insights.
Area of Science:
- Decision analysis
- Computational modeling
- Biostatistics
Background:
- Sensitivity analysis in complex decision models is challenging due to multiple uncertain input variables.
- Quantifying the impact of individual variables on decisions is crucial for robust analysis.
Observation:
- A new analysis method combines stochastic simulation with logistic regression.
- The method assesses the dichotomous decision variable against all input variables.
Findings:
- This method provides a direct measure of input variable importance for decision outcomes.
- It was demonstrated on a clinical decision regarding anticoagulation for deep vein thrombosis in early pregnancy.
- Relative importance is calculated by comparing log odds changes from individual variable variations to total log odds changes.
Implications:
- The proposed method offers a simple yet powerful tool for quantitative insight into decision models.
- It aids in understanding the nuances and key drivers within complex decision-making processes.
- This approach can enhance the reliability and interpretability of model-based decisions in various fields.