Ullrika gave a talk on Bayesian Inference at the upcoming REACH Discussion Meeting on 27 October in Lund
Link to the talk and some files are here:
https://github.com/Ullrika/BayesianInference
My talk followed directly after a talk about Machine Learning which is a very nice
Ullrika travelled all the way to Liverpool to give a tutorial at the BayesDays. Nice event.
The material for the tutorial in on github: https://github.com/Ullrika/TBDL2017
The seminar was led by the participants and gently facilitated by Ullrika. We went go through the book chapter by chapter. It was a very useful and inspiring day. Thank you all for coming. The old Bishop house in Lund is an excellent place for smaller workshops.
This is a seminar supported by the strategic research environment BECC and arranged by the BECC action group “Evidence relying on simulation models and expert judgment”
This book is intended for people in government, regulatory agencies and business who routinely make decisions and who rely on scientific and technical expertise. So-called evidence-based decision-making has become more popular over the last decade, but often the evidence we need for these decisions is unavailable. Time, money and the pressing nature of many decisions prevent us from collecting much of the information we need. In its place, decision-makers turn to experts to estimate facts or make predictions. The status of scientific and technical experts has evolved over the last 100 years or more to provide unprecedented opportunities for experts to influence decisions. The hidden risk is that scientists and other experts overreach, often with good intentions, placing more weight on the evidence they provide than is warranted. The tendency to overreach is pervasive and more significant than many scientists and decision-makers like to admit. Much of the evidence for these phenomena is drawn from well-established research on decision theory and cognitive psychology. This book documents the extent and importance of this issue, and then outlines a suite of simple, practical tools that will assist decision-makers to make better use of expert estimates and predictions. It provides the means to discriminate good advice from poor, and to help decision-makers to be reasonably and appropriately skeptical. The book promotes a change in attitude towards expert predictions and estimates such that they are treated with the same reverence as data, subjected to the same kinds of cross-examination and verification. By requiring a little discipline from their experts, decision-makers can avoid the most pervasive pitfalls of expert judgements and assure themselves of relatively reliable and accurate expert information.
Mark Burgman is Director of the Centre for Environmental Policy at Imperial College London and Editor-in-Chief of the journal Conservation Biology. Previously, he was Adrienne Clarke Chair of Botany at the University of Melbourne, Australia. He works on expert judgement, ecological modelling, conservation biology and risk assessment. He has written models for biosecurity, medicine regulation, marine fisheries, forestry, irrigation, electrical power utilities, mining, and national park planning.
The book
Trusting Judgements: How to Get the Best out of Experts by Mark Burgman. Cambridge: Cambridge University Press. 2015. doi:10.1017/CBO9781316282472
Ch1. What’s wrong with consulting experts?
Ch2. Kinds of uncertainty
Ch 3. What leads experts astray?
Ch 4. Dealing with individual experts
Ch 5. The wisdom of crowds revisited
Ch 6. Tips to get the best out of experts
The project Gaming for better decision making was one of the winners at the Climate-Kic Nordic Ideation Day in Aarhus August 23 2017.
Ullrika was there pitching it in front of a jury. Great day with a lot of positive feedback and useful connections.
Gaming for better decisions under uncertainty is a project with the aim to develop a computer game that motivates people to learn about uncertainty analysis and decision making in an entertaining way. The game will explain the benefits of expressing uncertainty when making predictions and possible ways to make decisions under uncertainty.
The game will include prediction and decision problems from daily life that players can relate to, and a fictional game scenario that involves decision making under climate uncertainty, where there is a need to consider the balance between social, economic, and environmental impacts. A case-study on climate decisions under uncertainty will be developed on flood protection in the municipality of Vejle, Denmark. We will benefit from previous experience in serious gaming at Aarhus University.
By the end of 2017 we will have an operational game prototype.
The game is expected to result in
– a greater interest to learn more about uncertainty among experts and decision makers
– a wider use of subjective probability for uncertainty
– a tool to initiate discussions on how to adapt uncertainty analysis to your decision problem
Partners: Lund University, DTU, Aarhus University, Vejle municipality
People (to be updated):
Ullrika Sahlin, Associate Professor, Lund University, Sweden. Uncertainty analysis in risk and environmental impact assessment.
Anthony O’Hagan Emeritus Professor, the University of Sheffield, UK. Eliciting expert knowledge using subjective probability; e-learning to train experts to express uncertainty accurately and rigorously.
Igor Linkov, Adjunct Professor, Carnegie Mellon University. Expert on decision making with applications on climate adaption and development of serious games of decision making.
Igor Kozine, Senior Researcher and Miroslava Tegeltija, PhD Student. Technical University of Denmark (Danmarks Tekniske Universitet). Quantitative and semi-quantitative approaches to uncertainty representation.
Matthias C. M. Troffaes. Expert on decision making under uncertainty with imprecise probability; applications in engineering and environmental sciences; avid gamer. Associate Professor (Reader), Durham University, UK.
External reviewer:
EFSA, Assessment and Methodological Support Unit, Olaf Mosbach-Schulz. Implementing uncertainty analysis in food and feed risk assessment. Training programs on probabilistic judgements for external experts.
The Nordic Chapter of the Society for Risk Analysis (SRA) invites abstracts to the 3rd conference which will be held at Aalto University, Espoo, Finland, November 2-3, 2017. (20 minutes’ bus trip from Helsinki).
The theme of the conference “Risk and Security” highlights the role risk plays in relation to security.
If you attend this conference you can listen to talks on risk and security from different disciplinary perspectives covering maths to social science
See more on the conference web
Register before October 17th
A Bayes@ mini-conference is a (more or less) local one day event bring together people working with or interested in Bayesian methods. Bayes@ events aims at being accessible to people with little experience of Bayesian methods while still being relevant to experienced practitioners. The focus is on how Bayesian methods are used in research and industry, what the advantages and challenges are with using Bayesian methods, and how Bayesian methods can be used and taught in a better way.
The first Bayes@ event was organized in Lund, Sweden by Rasmus Bååth and Ullrika Sahlin in 2014 when they noticed that, while many people use Bayesian methods at the university, they were spread all over campus. Rasmus and Ullrika thought: Why not meet up and share experiences and tips in a one day mini-conference? And so they did.
Read more on the github page bayesat.github.io/
Contact Ullrika or Rasmus if you are interested in arranging a Bayes@ conference. Have fun : )
Last Friday April 28th I gave my docenture lecture for Environmental Science at Lund University. I talked about the importance of considering extreme and rare events in environmental management. It was fun. Thanks all for coming. My presentation is here
Abstract: Have you ever missed a train or got stunned by a surprise you did not expect to happen? In retrospect, could you have done something to prevent bad things from happening and still get the good bits? These questions are valid, not only for our daily decisions, but also for decisions on how to manage the natural environment for conservation, safety or sustainability. To help out, there are theories for considering the extreme, rare or unlikely in decision making, with applications to the management of environmental systems.
Guess your time in Lundaloppet 2017 and express your uncertainty in your running time in which way you want!
After the race on May 13th in Lund and the actual running times are known, we will evaluate the performances of all predictions. The results will be presented in June, where we will hand out several awards.
Make your forecast here 25 people have guessed their running time. Thank you all for participating.
The challenge is open for anyone
The purposes with the challenge are to highlight what it means to forecast with uncertainty and collect examples of how people prefer to express uncertainty. Any way to express uncertainty is allowed, ranging from qualitative to quantitative, using probabilistic and non-probabilistic descriptions.
The participants are asked to guess their own running time and are therefore basing the forecast on best available (expert’s) knowledge. The nature of the question force them to think about uncertainty as their own, which demonstrate the subjective nature of uncertainty. Finally, participants are asked to rate their confidence in their forecasts, which is a more frequently occurring way to judge strength in predictive knowledge. 
The Lundaloppet predictive challenge was previously been given 2014 with great success.
This challenge is organized by the research group UnEviL (Uncertainty and Evidence Lab) at the Centre of Environmental and Climate research and is an activity of the BECC action group: Evidence relying on simulation models and expert judgment.
For more info contact ullrika.sahlin@cec.lu.se
Uncertainty and Evidence Lab is the name of a new research group at Lund University. The group is led by Ullrika Sahlin at the Centre of Environmental and Climate Research.
This research group focus on the management of uncertainty and evidence in risk analysis, evidence synthesis, decision analysis and, never the least, predictive sciences. This include discussing and applying principles and methods for learning, forecasting and decision analysis under different strengths in knowledge. We seek to apply the principles and methods on conceptual and actual problems and we are not limited to a specific application.
The research group, ongoing activities and output are presented at Lund University research portal
The program for Bayes@Lund2017 is now ready
Follow us at #bayeslund17 on twitter
We have uploaded videos of some of the talks on YouTube. Unfortunately not all talks have been recorded because of technical issues.
Richard McElreath — Bayesian Statistics without Frequentist Language: https://youtu.be/yakg94HyWdE
Martin Stjernman — Joint species modelling. Beautiful in theory, tricky in practice: https://youtu.be/KL9iKx0UKco
Shravan Vasishth — Finite mixture modeling: a case study involving retrieval processes in sentence comprehension: https://youtu.be/02YFuxJmyEI
Mark Andrews –Teaching Bayesian methods to social scientists: https://youtu.be/_gAg9UG9RCA
Stefan Wiens — Making the most of your ANOVAs: From NHST to Bayesian analyses: https://youtu.be/umjX4rnxAm0
Judith Bütepage — Learning to make decisions under uncertainty: https://youtu.be/1nz_96gxFvw
Ullrika Sahlin — Using expert’s knowledge in Bayesian analysis: link to presentation no video unfortunately Sahlin_BayesatLund2017
We start in room MA4, Maths building Annex, Sölvegatan 20. We also have a booklet of the abstracts programBayes@Lund2017 maps and tips.
[for the tutorial on 19 April go here]
08.30-9.00 Registration
09.00-9.10 Welcome and an overview of Bayesian activities in Lund: Umberto Picchini and Ullrika Sahlin
9.10-10.05 Keynote talk Darren Wilkinson: Hierarchical modelling of genetic interaction in budding yeast
10.05-10.30 coffee break
Bayesian Analysis I
10.30-10.55 Stefan Wiens, Making the most of your ANOVAs: From NHST to Bayesian analyses
10.55-11.20 Martin Stjernman, Joint species modelling — beautiful in theory, tricky in practice
11.20-11.45 Shravan Vasishth, Finite mixture modeling: a case study involving retrieval processes in sentence comprehension
11.45-13.05 Lunch break
13.05-14.00 Keynote talk Richard McElreath: Understanding Bayesian statistics without frequentist language
Decisions and Teaching
14.00-14.25 Judith Butepage: Learning to make decisions under uncertainty
14.25-14.50 Mark Andrews: Teaching Bayesian Data Analysis to Social Scientists
14.50-15.10 coffee break
Parallel Sessions
Bayesian Analysis II (room MA7)
15.10-15.35 Thomas Hamelrick: Potentials of mean force for protein structure prediction: from hack to math
15.35-16.00 Junpeng Lao: Statistical Inferences of Eye movement data using Bayesian smoothing
Teaching Bayes (room MA6)
15.10-15.35 Richard Torkar: Convincing researchers to transition to Bayesian statistics – the case of software engineering
15.35-16.00 Bertil Wegmann: Experiences from teaching Bayesian inference to students familiar with frequentist statistics
all back in room MA7 for the final session
Bayesian Analysis III (room MA7)
16.05-16.30 Erik Lindström: Multilevel Monte Carlo methods for inference in multivariate diffusions
16.30-16.55 Ullrika Sahlin: Using expert’s knowledge in Bayesian analysis
Funding from the research schools BECC and COMPUTE is greatly appreciated
Comments