Workshop

The aim of the workshop is to introduce the concept of mediated modelling in the one-health context. We want to show how this method might help to cope with the complexity of zoonotic diseases. The participants will learn about
System dynamics thinking – the basis of mediated modelling,
How to interactively involve affected stakeholders in the modelling process,
How mediated modelling might allow policymakers and other stakeholders to see the consequences of their actions over longer time scales.

Understand the problems associated with inferential modelling of high dimensional, wide data including inappropriate covariate selection, coefficient inflation and over fitting.
Content / structure
Outline why conventional methods of model selection (e.g. a univariable filter followed by stepwise selection) perform poorly with high dimensional data.
Describe the principles and issues of overfitting including when and why it is likely to occur and how to detect it
Describe the impact of overfitting on inference
 
Understand different approaches to overcome problems when modelling high dimensional, wide data including regularisation and altering the variance bias trade off. Understand the strengths and weaknesses of different approaches.
Content / structure
Outline the principle of variance-bias trade off and how this can help to overcome issues of over fitting when the number covariates (p) is greater than the number of observations (n)
Describe the principles of regularisation and discuss details of methods including lasso, elastic net, MCP and modified Bayesian Information Criterion
Describe advantages and disadvantages of different approaches
 
Learn to implement basic regularisation techniques on high dimensional data including elastic net, minimax convex penalty (MCP), and modified BIC using 'R'.
Content / structure
Methods of regularised regression will be presented as examples and delegates will be able to fit models on their own laptops using data and code provided.
Model interpretation and assessment of fit will be discussed
Comparison of results between regularised methods and conventional model selection based on AIC will be discussed 
 
Understand the concept of covariate selection stability and learn how to implement it using regularised models in ‘R’.
Content / structure
Define covariate selection stability in regression models
Outline the theoretical principles of selection stability and describe its usefulness for robust inference.
Describe methods to illustrate selection stability graphically.
Practical implementation; delegates will have the opportunity to evaluate selection stability using simulated datasets with known underlying casual variables (including random effects models) using code provided in R software
 
Understand the concept and importance of multiple method triangulation and learn how to implement this using ‘R’.
Content / structure
Outline the principles of multiple method triangulation and its relevance in terms of reproducible science
Demonstrate the concept using triangulation of a variety of methods of regularised regression
Practical implementation; delegates will have the opportunity to conduct multiple method triangulation using code provided in R software
 

The aim of the workshop is to introduce participants to using phylodynamic approaches to study virus outbreaks.
 
In the first part of the workshop, we will introduce participants to the key theoretical concepts that underpin the use of phylodynamic approaches. We will then support participants to work through two tutorials on their own laptops, giving them practical experience of applying these tools to reconstruct to reconstruct viral epidemiological dynamics.
 
The first tutorials will focus on identifying dates when outbreaks may have originated, and seasonal change in virus outbreak size. Following a short break, we will then introduce approaches for estimating virus geographical spread and detecting covariates associated with spread in a second tutorial.
 
Whilst the tutorials will focus on curated test data provided by the facilitators, we will briefly discuss what features make a virus genomic dataset ideal for analysis using these methods so that participants feel confident in planning their own studies following the workshop. Participants will be encouraged to discuss strategies for analysing their own data with the facilitators during a mid-workshop break.
 

Duration 9.00-12.00 hr = 180 min (rough planning 165 min)
 
Short introduction to each other: ~ 15 min
Cross the line principle (so active), based on few questions around study design and bias (5 min/question ?).
 
Plenary ppt: ~ 20 min
Short introduction on the 3 basis design issues (see objectives).
 
Working in small subgroups on a study design: ~ 30 min
We will provide 5-6 veterinary relevant research questions (in different species/systems) and let subgroups (2-3 persons) work out the most appropriate design & handling of bias for a chosen research question. In any case the same research question will be assigned to pairs of subgroups. Both facilitators will roam around to support subgroup discussions.
 
Break: ~ 15 min
 
Working in pairs of subgroups: ~ 40 min
Both subgroups (4-6 persons) provide feedback to each other’s design proposal for the same research question. Based on this group discussion they will define 2-3 learning points for plenary closure. Facilitators will roam again.
 
Plenary closure: ~ 45 min
Subgroup pairs present 2-3 learning points to all.
Plenary discussion about those learning points.

Livestock and aquaculture are a critical part of economies contributing to livelihoods, businesses and employment. They also provide opportunities for exports and food in a national context for the health and wellbeing of the billions of consumers. Diseases and health problems constrain the livestock and aquatic sectors in their ability to contribute to the social, economic and cultural outcomes of society. Understanding how to effectively and efficiently address these impacts requires systematic data collection and analysis in order to quantify and understand the burden of animal diseases. The GBADs programme is addressing this problem to provide an understanding of resource allocation for animal health and welfare. It will provide opportunities for decision makers to critically assess the balance of net production losses and expenditure on critical health issues that are limiting animal productivity, welfare, compromising human health (i.e. zoonoses and antimicrobial resistance) and contributing to inefficient use of land and water resources. GBADs is funded by the Bill and Melinda Gates Foundation and The UK’s Foreign, Commonwealth and Development Office and led by the University of Liverpool with the World Organisation for Animal Health (WOAH).

The workshop will consist of a combination of powerpoint lectures and hands-on exercises to get acquainted with generic risk assessment tools for disease incursion. In the first part of the workshop, the participants will be introduced to the objectives of generic risk assessment, approaches for generic risk assessment as well as data requirements and communication of results. Two generic risk assessment tools recently developed at Wageningen Bioveterinary Research1 will be presented in more detail, explaining the risk steps and pathways included in these tools and the algorithms used to evaluate the risk. Participants will have the opportunity to explore both tools in more detail by working on an assignment for each tool, using the online user interfaces that were developed.
 
1 Rapid risk assessment tool (RRAT) for emerging and re-emerging diseases; Model for INTegrated RISK assessment of vector-borne diseases (MINTRISK)