Grid View

Tutorial on quantifying epistemic uncertainty by probability

This is a 3 hour tutorial where I (Ullrika Sahlin) talk about why quantification of epistemic uncertainty by probability is important and useful to master. I present two methods – Expert Knowledge Eliciation and Bayesian Analysis.

We will go through hands on examples on quantifying uncertainty in assessment inputs as well as in assessment output.

All examples will be done using code from the open source program R using the participants own laptops. It is possible to just sit in and interact without running the code.

The purpose of the tutorial is to spark interest in using probability as a measure of epistemic uncertainty. Instead of going through the maths behind it, I will walk you through hands on examples which demonstrate that it is not difficult to do Bayesian analysis in practise. We use the SHELF R-package for EKE and rjags and JAGS for MCMC sampling.

This tutorial was given in Lund 10th of October 2018

I will give it in Stavanger on November 7th between 16 – 19 in Kjølv Egelands hus, Room D301 . The tutorial is part of a conference in risk https://www.uis.no/research-and-phd-studies/research-centres/seros-centre-for-risk-management-and-societal-safety/risk2018/ Contact me if you are interested in attending the tutorial ullrika.sahlin [at] cec.lu.se or sign up for it when you register to the Conference.

The material for the tutorial is available at GitHub. Note – work in progress.

https://github.com/Ullrika/quantifying_uncertainty_by_probability

No one wants to miss the last bus (giphy.com). We do a risk analysis of missing the last bus, without and with epistemic uncertainty.
The side that comes up when throwing a dice is subject to aleatory uncertainty. Not knowing the number of sides on a dice is epistemic uncertainty. You can reduce epistemic uncertainty by observing which dice it is, but you cannot reduce aleatory uncertainty by throwing the dice (not even if done many times).
I asked the participants to express their uncertainty in the likelihood that Sweden will have the Winter Olympic Games 2026.

 

 

October 13, 2018

This entry was posted in

risk concept teaching

Comments

0 Comments Leave a comment

Are you uncertain?

This is a question we asked many times last Saturday when the Uncertainty Show was given for the first time.

The show relies on people coming up on stage trying to be uncertain.

They compete against a computer, theory or others.

A lot of people tried our computer games about uncertainty.

Ivette, Maria, Ullrika and Rasmus – happy after making so many people uncertain.

September 17, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

Open PhD position in decision theory and environmental management

The research Group Uncertainty and Evidence Lab at the Centre of Environmental and Climate Research are looking for a suitable candidate for a new PhD position.

The aim of the PhD project will to study the importance of uncertainty in environmental management. This includes the treatment of uncertainty into scientific assessments and considering in what way decision makers respond to uncertainty. The impact of uncertainty will be studied by ecological-economic modelling focusing in solving environmental decision problems. The project includes a case-study on geese management in agricultural landscapes, representing a spatial-temporal ecological-economic system with several and different types of agents, both human (farmers) as well as ecological agents (geese).

Read more here about the position and how to apply.

Last day to apply is October 1st 2018.

August 30, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

Premiere of the mini-show “The art of being uncertain” at Lund University Science Festival 15th of sept 2018

The art of being uncertain is the name of our new mini-show, which has its premiere on September the 15th at the Lund University Science Festival.

This years logo for the uncertainty show

PLACE: ASTRONOMIHUSET, SÖLVEGATAN 27 11.00–17:00

Through games and challenges, you will learn more about different ways to express uncertainty and how uncertainty influences us as human beings.

This is an attempt to reach out with the fascinating stuff and important things to know about uncertainty. The people behind the show are Ullrika Sahlin and Rasmus Bååth.

The program for the Science festival Naturvetarstråket contains a lot of activites for a curious mind.

There will be possibilities to play games about uncertainty. These games are available on Android from google app store. The apps are found by searching for UncertainGames. There are five English (Quiztimate, BeanGuesser, ProbabilityGuess, Probability Bee and FrequencyGuesser) and one Swedish game (Gisso).

The games are also possible to download for windows and mac.
https://uncertaingames.itch.io/

 

August 29, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

4th Nordic Chapter Risk conference in Norway Nov 8-9

Exploring the risk, safety, security and resilience nexus

Risk, safety, security and resilience analysis and management are founded on similar ideas and perspectives: avoiding or reducing the negative consequences of hazardous events and threats, exploring system weaknesses, and identifying policies and actions to mitigate or resolve such weaknesses. However, current research and practices in these areas are to a large extent disconnected and rooted in separated ‘schools’, providing guidance in different directions.  For instance, many researchers, practitioners and organisations have called for a shift from risk to resilience. In what ways, however, may risk and resilience represent conflicting perspectives? Or could it be argued that resilience analysis is a main strategy within risk management? Efforts have been made to develop more unified approaches and methods, but how best to proceed to change the current situation?

The conference theme addresses the risk, safety, security and resilience nexus. We seek contributions that can provide new insights on this topic, in particular theories, principles, approaches and methods that can enhance the development of more integrated perspectives.

The 4th SRA Nordic Chapter conference aims to explore the status quo and the future directions of risk research. The conference wishes to harness its broad, interdisciplinary expertise by gathering researchers from academic institutions and practitioners from the risk, safety, security and resilience communities. We intend to build networks and discuss the most pressing issues within this area of research. The conference gathers academics, policy makers, and other professionals interested in risk research, policy, and practice.

The conference is organised in collaboration with ESRA Norway. We are pleased to announce that all participants at this conference will automatically become members of SRA worldwide and SRA Europe!

For additional nformation about the conference (including instructions for submitting abstracts), please see the conference website.

July 13, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

Opportunities to dive into uncertainty

Uncertainty is more important than ever. Not only are problems of today more complex and urgent, but there is also a need to combine different types of knowledge and weight things together. In order to give scientific advice we must train our ability to state what we dont know.

The uncertainty and evidence lab are working on taking these type of questions seriously.

Games as measuring instruments

The games to explain what it means to express uncertainty have been developed further into not only amusement and teaching but also as measurement instruments of peoples ability to express their own uncertainty.

We have started to explore the possibility to makes the games available for phones to improve testing and their use for research. Two versions are on Google app store: Probability bee and quiztimate. So far there are only solutions for Android up and running.

Take part of the discussion

Many risk Conferences bring up the issue of uncertainty. Such as this one. Will I see you there?

Keep posted

 

July 13, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

Working from home

We are approaching the time of the year to writing applications for the Swedish research councils VR and FORMAS. We all have our tricks to get the opportunity to focus and bring those ideas on a piece of paper (literary). Today I skip commuting and I work from home. Started with a nice walk after leaving my son at the school bus.The sun was rising in a fog and I took the walk around the woods.

Back, I have the opportunity to use my nice and sunny room.

Countryside is both good and bad. Nature is close, but one travels a lot to activities, friends and work. In May, the permanent telephone will be removed for good (and I live in a very populated part of Sweden). Unless we dig down a cable, the only option is mobile internet. But first I have to finish those research applications.

 

March 6, 2018

This entry was posted in

research life

Comments

0 Comments Leave a comment

Reflections on stakeholder interaction

Today I gave a talk at the CEC Fellows meeting on Lessons learnt from interacting with stakeholders. I shared reflections on three interactions I have been involved with:

  • Honey bees: bee health, pesticides and pollination
  • Invasive species: risk assessment and signal crayfish
  • Uncertainty in scientific assessments: understand and quantify!

The slides of my talk are found here: Sahlin_Lessons learnt from interacting with stakeholders_public

Ways to reach a systematic and science-based approach to stakeholder interaction may be to provide opportunities to learn from experience of interaction, help you to identify what type of interaction you want to work with and how much, and stimulate scientific discussion on interaction (science dialogue, communication, knowledge production, …).

 

January 29, 2018

This entry was posted in

info meetings

Comments

0 Comments Leave a comment

Master thesis: Can drones improve detection and treatment of potato late blight?

Early detection and recommendation of higher precision in treatment of late blight in potato crops by using drone photos is the title of a Master Thesis by Anna Lindéus at the Risk Master program at Lund University (LTH). It has been developed in a collaboration with the Centre of Environmental and Climate research and Vultus. The supervisors have been Ullrika Sahlin (CEC) and Abdalla Eltayeb (Vultus).

The thesis was presented Friday 26th at 10.00 in V:Brand 3th floor, V-building, LTH, Lund.

Photo: A Lindeus

Description in Swedish:

Förekomsten av bladmögel hos potatis har varit och är ett globalt problem. Bladmögel orsakade t.ex. den förödande hungersnöden på Irland i mitten av 1800-talet. I dag besprutas potatisodlingar med stora mängder bekämpningsmedel, i form av fungicider, i förebyggande syfte för att motverka bladmögel.

Genom att försöka undvika onödiga och överflödiga fungicidbehandlingar, kommer det att minska kostnaderna för brukarna samt gynna miljön. Forskning pågår för att försöka optimera användningen av fungicider. Det finns redan flera beslutsstödsystem som ger rekommendationer gällande hantering av bladmögel. Dessa beslutsstödsystem använder modeller som försöker förutspå risken för mögelangrepp, som framförallt är baserade på väderdata. Fjärranalys och drönarfoto används för precisionsodling, vilket är ett sätt att minska kostnader och miljöpåverkan.

The influence diagram of the decisions to use drones and to spray with fungicides.

I detta arbete har en riskmodell utvecklats för att förutse risken för bladmögel, baserad på väderdata och drönarfoto. Riskmodellen har integrerats i en beslutsanalys för att undersöka om det är kostnadseffektivt att använda drönare för att upptäcka mögel. En känslighetsanalys med fiktiva värden för kostander visar att det man vinner i känslighet mot huruvida det finns mögel eller ej är svårt att uppväga mot möjliga förluster i potatisskörden, särskilt när kostand för drönarfoton är hög. Det skulle kunna gå att hitta strategier där drönare ger mervärde om de kan minska kostnad för applicering av fungicider.

Vi planerar ta arbetet ett steg till och göra en något mer avancerad modell som ser på beslut och utveckling av mögel vid flera tillfällen och som bygger på mer data från fältförsök och drönarfoton.

January 23, 2018

This entry was posted in

Uncategorized

Comments

0 Comments Leave a comment

Bayesian Analysis and Decision Theory Spring 2018

A graduate course in Bayesian Analysis and Decision Theory

This course is hosted by the research school ClimBEco and is open for anyone.

Registration is closed

The course has three physical meetings during spring 2018: 13-14 March, 20-21 March, 10-11 April badt_2018_schedule_15feb

This course is worth 5 credits –  Course Syllabus

Bayesian methods are used for inference, modelling of complex data, model calibration, and integration of multiple sources of information, combination of data with expert knowledge, risk and decision analysis. This is a course which mix theory and literature seminars with hands on exercises on Bayesian analysis to learn, predict, quantify uncertainty and make decisions.

The course will be open for anyone who has basic knowledge in probability theory or statistics. Experience in R will be an advantage for the hands on exercises. Methods will be presented in settings where they are applied using simple examples from the field of environmental and risk management. The purpose with the applied focus is to give students basic skills to use these methods, understand what they can be used for, and stimulate the student’s curiosity in learning more about them at a more foundational level.

Upon completion of the course, the student shall:

  • Be able to give an account of the principles behind Bayesian decision analysis,
  • Be able to give an account of the principles behind Bayesian and hierarchical modelling,
  • Be able to give examples of principles to quantify and treat uncertainty in quantitative assessments,
  • Be able to give an account of science theoretic arguments behind principles to quantify and treat uncertainty in knowledge production and decision making

The course consists of lectures, exercises, literature seminars and an individual project covered by three physical meetings and self studies. Case studies will be used cutting across the different parts.

Part I (March 13-14)

Ullrika Sahlin and Niklas Vareman

  • Introduction to decision analysis
  • Bayesian Belief Networks
  • The philosophy of uncertainty

Part II (March 20-21)

Johan Lindström

  • Bayesian Hierarchical Modelling in JAGS and INLA.

Part III (April 10-11)

Ullrika Sahlin, Niklas Vareman and Inari Helle (University of Helsinki)

  • Bayesian decision analysis
  • Decision making robust to uncertainty – in theory and practice
  • Limitations and extensions of Bayesian decision analysis

Part IV. Individual project

The purpose with the individual project is to give the students an opportunity to deepen or apply course content on something that are in their own interest. It can be purely theoretical or more hands-on modelling project. A suggestion is to do a decision analysis where one need to integrate data and expert judgment.


The schedule will be specified so that it is possible to travel back and forth from nearby cities over the day starting at 10 and ending at 15, with a possibility of Q&A between 15 and 16 the first day of every two day sessions.

In addition, course participants are invited to participate for free in the one day conference Bayes@Lund taking place in Lund on April 12th.

This course was given as a summer school 2015 – I have pasted the blog post we had back then

First week completed of the summer school Bayesian Analysis and Decision Theory

For questions contact me Ullrika — ullrika.sahlin [at] cec.lu.se

January 12, 2018

This entry was posted in

students teaching

Comments

0 Comments Leave a comment

Newer Posts Older Posts