Introduction to assessment performance Mikko Pohjola, THL.

Introduction to assessment performanceMikko Pohjola, THL

Contents

• Concepts

• General framework

• Common perspectives (& examples)• Quality assurance/quality control• Uncertainty• Model performance

• Properties of good assessment

• Summary

Concepts & rationale

• Some basic concepts:• Performance = goodness!• Assessment, Management• Model• Process (making/using), Product• Output, Outcome• Assessor, Decision/Policy maker, Stakeholder• Participant, User

Concepts & rationale

• Why evaluation of assessment performance is important?

• Efficient use of resources?• Value of work done?• Importance/meaning of information?• Implications of information?• Actual impacts of information?• …• …because funder, customer, user, boss, peer,

stakeholder etc. wants/needs to know!

Roles and interests

Experts Data quality, analysis procedure, coherence, comprehensiveness, …

Funders Relevance, efficiency, timeliness, importance, …

Users (DM) Understandability, reliability (of source), acceptance, practicality, …

Interested (SH) (same as DM, but different perspective)

General RA/RM framework

• Process, product, use

Assessment Use

Assessmentprocess

Assessmentproduct

Decision making

Knowledgeneed

Productrequirement

Processrequirement

Common perspectives & examples

• Quality assurance/quality control• Focus on assessment process• An “engineering” perspective

• Uncertainty• Focus on assessment output• A scientists perspective???

• Model performance• Focus on modelling and model

• Combines QA/QC and uncertainty perspectives• A modellers perspective

Quality assurance/quality control

• Principle:• Good process guarantees good outputs/outcomes!

• Question:• How should an assessment process be conducted?

• Examples:• Ten steps by Jakeman et al.(2006)• IDEA framework (Briggs, 2008)• (Over)appreciation of randomized controlled trials

(RCT’s)

Ten iterative steps in development and evaluation of environmental models

Jakeman et al.: Ten iterative steps in development and evaluation of environmental models. Environmental Modelling & Software Issue 5, May 2006, Pages 602-614

IDEA framework (INTARESE)

Briggs: A framework for integrated environmental health impact assessment of systemic risks. Environmental Health 2008, 7:61.

Uncertainty

• Principle:• Performance is an intrinsic property of an information

product!

• Question:• How good is the answer provided by the

assessment?

Uncertainty

• Examples:• Statistical uncertainty analysis

• Mean, variance, confidence limits, distributions, …• Cf. D. Lindley: Philosophy of Statistics, 2000

• Sources of uncertainty• E.g. model, parameter & scenario uncertainty (as

applied e.g. by the U.S.EPA)• Extensive approaches

• E.g. inclusion of qualitative aspects, sources of uncertainty as in NUSAP (www.nusap.net)

NUSAP

• N: numeral

• U: unit

• S: spread

• A: assessment (qualitative judgment)

• P: pedigree (historical path leading to result)

NUSAP - pedigree

Jeroen van der Sluijs: NUSAP- some examples. Presentation. Available: http://tinyurl.com/5uwln2r

Model performance

• Principle:• The model is the essence of the assessment!

• Question:• How good is the model?

• Examples:• Verification, validation, (reliability, usability, …)• Outcome-oriented approach by Matthews et al. 2011

Outcome-oriented modelling approach

Matthews et al.: Raising the bar? – The challenges of evaluating the outcomes of environmental modelling and software. Environmental Modelling & Software, March 2011, Pages 247-257.

Summary of common perspectives

• Assessment process and product addressed in many ways

• Use of results mostly not considered• The link between outputs and outcomes (cf. Matthews

et al. 2011)• Evaluation often a separate process

• Expert processes of making assessments and using their results

• Expert processes of evaluating performance

• Alternative perspectives?

Properties of good assessment

Properties of good assessment

• Ex post (after assessment) evaluation

• Ex ante (before/during assessment) evaluation• Guidance of design and execution

• Links process and output with use• Thereby also linking them to outcomes

Example: what makes a good hammer?

Example: what makes a good hammer?

• How is the hammer made? By whom?

• What properties does the hammer have?

• What do you want to do with the hammer?

• How does the hammer help you do it?

Summary

• Consideration of (intended) use is essential• Consideration of process and product in light of use

• Consider the instrumental value of information• Cf. absolute value (a common science view)• Cf. Ad hoc solutions (a common practice view)• Contextuality, situatedness, practicality, …• In policy-support information is a tool (a means to an

end)• A model is a tool for producing information

• How does this relate to the previous lectures about DA and the DA study plan exercise?