Stakeholder Preference Modeling with Probabilistic Inversion Roger M. Cooke Resources for the Future Dept Math, TU Delft June 16, 2011.

Slides:



Advertisements
Similar presentations
Extending the external costs framework Prof. Anil Markandya University of Bath External costs of energy and their internalisation in Europe Dialogue with.
Advertisements

The Simple Linear Regression Model Specification and Estimation Hill et al Chs 3 and 4.
National Institute for Public Health and the Environment Priority setting of emerging zoonoses Marieta Braks, Ph.D. and Prof. A. Havelaar, M. Toutenel.
Discrete Choice Modeling William Greene Stern School of Business IFS at UCL February 11-13, 2004
DETC06: Uncertainty Workshop; Evidence & Possibility Theories Evidence and Possibility Theories in Engineering Design Zissimos P. Mourelatos Mechanical.
Discrete Choice Model of Bidder Behavior in Sponsored Search Quang Duong University of Michigan Sebastien Lahaie
Using expert judgments to explore robust alternatives for forest management under climate change By Tim McDaniels University of British Columbia and CEDM.
2 – In previous chapters: – We could design an optimal classifier if we knew the prior probabilities P(wi) and the class- conditional probabilities P(x|wi)
Hydrogen Production Decisions: Decision Making in View of Differing Stakeholder Preferences Elvin Yuzugullu Doctoral Candidate The George Washington University.
TNO orbit computation: analysing the observed population Jenni Virtanen Observatory, University of Helsinki Workshop on Transneptunian objects - Dynamical.
©GoldSim Technology Group LLC., 2004 Probabilistic Simulation “Uncertainty is a sign of humility, and humility is just the ability or the willingness to.
What determines student satisfaction with university subjects? A choice-based approach Twan Huybers, Jordan Louviere and Towhidul Islam Seminar, Institute.
Certainty Equivalent and Stochastic Preferences June 2006 FUR 2006, Rome Pavlo Blavatskyy Wolfgang Köhler IEW, University of Zürich.
Choice Modeling Externalities: A Conjoint Analysis of Transportation Fuel Preferences Matthew Winden and T.C. Haab, Ph.D. Agricultural, Environmental,
Hilbert Space Embeddings of Hidden Markov Models Le Song, Byron Boots, Sajid Siddiqi, Geoff Gordon and Alex Smola 1.
CS 589 Information Risk Management 6 February 2007.
A Thousand Flowers, A Thousand Weeds: New Challenges to the Rationality of Risk* Eugene A. Rosa Professor of Sociology Edward R. Meyer Professor of Natural.
1 Empirical Similarity and Objective Probabilities Joint works of subsets of A. Billot, G. Gayer, I. Gilboa, O. Lieberman, A. Postlewaite, D. Samet, D.
Department of Engineering and Public Policy Carnegie Mellon University Integrated Assessment of Particulate Matter Exposure and Health Impacts Sonia Yeh.
1 Stochastic Dominance Scott Matthews Courses: /
Modelling Cardinal Utilities from Ordinal Utility data: An exploratory analysis Peter Gilks, Chris McCabe, John Brazier, Aki Tsuchiya, Josh Solomon.
Executive Manager Decision Making and Policy Planning, typically with many goals Sometimes even > 1 decision maker (Game Theory, Group Decisions) Linear.
Using ranking and DCE data to value health states on the QALY scale using conventional and Bayesian methods Theresa Cain.
Lecture II-2: Probability Review
Standard error of estimate & Confidence interval.
Slide 1 Tutorial: Optimal Learning in the Laboratory Sciences Working with nonlinear belief models December 10, 2014 Warren B. Powell Kris Reyes Si Chen.
«Enhance of ship safety based on maintenance strategies by applying of Analytic Hierarchy Process» DAGKINIS IOANNIS, Dr. NIKITAKOS NIKITAS University of.
Fundamentals of Statistical Analysis DR. SUREJ P JOHN.
Marko Tainio, marko.tainio[at]thl.fi Modeling and Monte Carlo simulation Marko Tainio Decision analysis and Risk Management course in Kuopio
The Development of Decision Analysis Jason R. W. Merrick Based on Smith and von Winterfeldt (2004). Decision Analysis in Management Science. Management.
STA291 Statistical Methods Lecture 16. Lecture 15 Review Assume that a school district has 10,000 6th graders. In this district, the average weight of.
Paul Bakker – Social Impact Squared
Simon Thornley Meta-analysis: pooling study results.
Probabilistic Robotics Bayes Filter Implementations Gaussian filters.
ECE 8443 – Pattern Recognition LECTURE 07: MAXIMUM LIKELIHOOD AND BAYESIAN ESTIMATION Objectives: Class-Conditional Density The Multivariate Case General.
Science Advisory Panel Data Analysis Workgroup Paul Bukaveckas July
Continuous Variables Write message update equation as an expectation: Proposal distribution W t (x t ) for each node Samples define a random discretization.
Discrete Choice Modeling William Greene Stern School of Business New York University.
Using a Discrete Choice Experiment to Value the EQ-5D-5L in Canada Nick Bansback Assistant Professor School of Population and Public Health, University.
Risk Analysis & Modelling Lecture 2: Measuring Risk.
Discrete Choice Modeling William Greene Stern School of Business New York University.
Ch15: Decision Theory & Bayesian Inference 15.1: INTRO: We are back to some theoretical statistics: 1.Decision Theory –Make decisions in the presence of.
A Multi-Expert Scenario Analysis for Systematic Comparison of Expert Weighting Approaches * CEDM Annual Meeting Pittsburgh, PA May 20, 2012 Umit Guvenc,
Decision Making Under Uncertainty CMSC 471 – Spring 2014 Class #12– Thursday, March 6 R&N, Chapters , material from Lise Getoor, Jean-Claude.
Statistical inference Statistical inference Its application for health science research Bandit Thinkhamrop, Ph.D.(Statistics) Department of Biostatistics.
Stats Term Test 4 Solutions. c) d) An alternative solution is to use the probability mass function and.
- 1 - Outline Introduction to the Bayesian theory –Bayesian Probability –Bayes’ Rule –Bayesian Inference –Historical Note Coin trials example Bayes rule.
Generalized Point Based Value Iteration for Interactive POMDPs Prashant Doshi Dept. of Computer Science and AI Institute University of Georgia
Statistical Methods. 2 Concepts and Notations Sample unit – the basic landscape unit at which we wish to establish the presence/absence of the species.
6. Ordered Choice Models. Ordered Choices Ordered Discrete Outcomes E.g.: Taste test, credit rating, course grade, preference scale Underlying random.
Parameter Estimation. Statistics Probability specified inferred Steam engine pump “prediction” “estimation”
[Part 5] 1/43 Discrete Choice Modeling Ordered Choice Models Discrete Choice Modeling William Greene Stern School of Business New York University 0Introduction.
Confidence Intervals and Hypothesis Testing Mark Dancox Public Health Intelligence Course – Day 3.
Outline Historical note about Bayes’ rule Bayesian updating for probability density functions –Salary offer estimate Coin trials example Reading material:
AMBIGUITY 2 day short course on Expert Judgment Roger Cooke Resources for the Future Dept. Math, Delft Univ. of Technology April 15, UNCERTAINTY.
Centre for Transport Studies Modelling heterogeneity in decision making processes under uncertainty Xiang Liu and John Polak Centre for Transport Studies.
Uncertain Judgements: Eliciting experts’ probabilities Anthony O’Hagan et al 2006 Review by Samu Mäntyniemi.
Probabilistic Project Management for a Terrorist Planning a Dirty Bomb Attack on a Major US Port Workshop on Critical Infrastructure Protection June 5-7,
William Greene Stern School of Business New York University
Null Hypothesis Testing
Expert Judgment short course, NIA, 15,16 April, 2008
Computational models for imaging analyses
Microeconometric Modeling
Expert Judgment short course, NIA 15,16 April 2008
Statistical Thinking and Applications
Experrt Judgment short course, NIA 15,16 April 2008
CS639: Data Management for Data Science
Fig. 2 Meta-analysis: Country-specific effects on vote choice.
Chapter 3 Hernán & Robins Observational Studies
Presentation transcript:

Stakeholder Preference Modeling with Probabilistic Inversion Roger M. Cooke Resources for the Future Dept Math, TU Delft June 16, 2011

Foundations Health states Risks nano-enabled food

Expert Judgment for Uncertainty Quantification: PM 2.5 Uncertainty in Mortality Response to Airborne Fine Particulate Matter: Combining European Air Pollution Experts Jouni T. Tuomisto, Andrew Wilson, John S. Evans, Marko Tainio (RESS 2008)

Fundamental Theorem of Decision Theory For Rational Preference UNIQUE probability P which represents degree of belief: DegBel(France wins worldcup) > DegBel(Belgium wins worldcup)  P(F) > P(B) AND a Utility function, unique up to 0 and 1, that represents values: ($1000 if F, $0 else) > ($1000 if B, $0 else)  Exp’d Utility (($1000 if F, $0 else)) > Exp’d Utility (($9000 if B, $0 else)) BUT….

UNLIKE Expert Judgment: There is no Updating utilities on observations Convergence of utilities via Observations Empirical control on Utilities Community of ‘Utility Experts’ Rational consensus on Utilities

Why is Preference Modeling Impoverished? AHP MAUT MCDM ELECTRA REMBRANT OUTRANKING THURSTONE BRADLEY TERRY PROBIT LOGIT NESTED LOGIT PSYCH’L SCALING Validation ???

What means Validation? Goal = find ‘true Utility values’ for alternatives? Fools’s Errand

Condorcet’s Paradox of Majority Preference 1/3 prefer Mozart > Hayden > Bach 1/3 prefer Hayden > Bach > Mozart 1/3 prefer Bach > Mozart > Hayden THEN 2/3’s prefer Bach > Mozart Mozart > Hayden Hayden > Bach

What can we do? Random Utility Theory Each (rational) stakeholder has a utility function over alternatives  characterize population as distribution over utility functions

Probabilistic Inversion Domain: utility functions Of stakeholders Range : choices of stakeholders G maps utilities into choices Invert G at this distribution Observe Stakeholders Preferences

Used for stakeholder Preference Modeling: Risks of Nano enabled foods (Flari, WHO, CIS) Valuing impaired health states (Flari, FDA) Valuing fossil fuel policy options (RFF) Prioritizing ecosystem threats (NCEAS) Prioritizing zoonose threats (RIVM) Modeling wiring failure (Mazzuchi) Prioritizing vCVJ options (Aspinall Health Canada) UK Research Council (Aspinall) Aus. Univ. Fac Sci reviews (Aspinall).

steps 1.Get discrete choice data from stakeholders for choice alternatives A1,…An – “Which of (A,B) do you prefer” – “Rank your top 3 of (A, B, C, D, E, F)” 2.Find dist’n over utilities on [0,1] n which reproduces stakeholders preferences 3.If utility is function of covariates, validate out of sample.

Valuation of impaired Health states Flari et al 17 health states 6 criteria Each criteria has 3 values, described in narrative 19 Experts ranked 5 groups of 5 health states

First, find dist’n over utilities for the 17 Health States which recover Observed Frequencies of rankings (i.e. wo criteria)

Build MAUT model for HS utilities, based on the 6 criteria Each stakeholder has a weight vector that determines his/her preference Population of stakeholders = population of weights Characterize population based on all rankings involving at least 7 (30%) experts (= 28 rankings). Validate on remaining rankings (= 77 rankings)

Average weights

Preference dependence emerges from fitting

Average weights per group of 5 health states

Predict out-of-sample rankings First time in HISTORY that a multi attribute model has been WRONG!!!

Average of predictions vs Out-of- Sample observed rankings, Not SOOO bad

Risks of Nano-Enabled Foods Flari et al

Nano enabled food risks (VLARI) rank top and bottom 5

Correlation of criteria weights

Conclusion Stakeholder preference modeling is empirical science ‘preference for criteria’ inferred from data, not elicited (in) dependence in choices inferred from data, not assumed THANK YOU