Environmental Evidence Synthesis

Here we will post suggestions for master projects that in one way or another is using Bayesian modelling. We are looking for students eager to learn and develop. Feel free to contact us with your own ideas. Send an email to ullrika.sahlin at cec.lu.se!

Master project 1. An Hierarchical model of QSAR-integrated Hazard assessment

Master project 2. Building a Bayesian Belief Network for invasive species assessments

Master project 3. A Bayesian approach to cost-efficiency evaluation of agri-environmental schemes

Master project 4. Statistical testing without p-values using Bayesian Data Analysis

Master project 1. An Hierarchical model of QSAR-integrated Hazard assessment

The first one to go is a hierarchical model considering error in testing and non-testing information.
Supervisor is Ullrika Sahlin Centre of Environmental and Climate Research

Hazard assessments for environmental systems are based on toxicity values on several species representing different tropical levels and taxonomic groups in the ecological system. The Species Sensitivity Distribution (SSD) approach is commonly used to take variability between species into account when deriving the ecological relevant hazard concentration.

In the lack of appropriate testing information on toxicities, non-testing information can be used. Non-testing information can be derived from Quantitative Structure Activity Relationships (QSARs). Whereas testing information has associated errors, predictions from QSARs come with uncertainty due to both model error and error in underlying testing data used to train the model.

The aim of this project will be to build a hierarchical model of QSAR-informed hazards assessment that take error in data, model error in QSAR predictions and variability in species sensitivities into account. The model will be used to illustrate principles to establish evidence bases in environmental decisions.

The student will get the opportunity to learn Bayesian modelling in BUGs or JAGS.

Master project 2. Building a Bayesian Belief Network for invasive species assessments

The student brave enough to undertake this project will develop and evaluate the evidence bases of an existing Bayesian Belief Network to perform risk assessments of invasive species.

Master project 3. A Bayesian approach to cost-efficiency evaluation of agri-environmental schemes

We have access to a lot of empirical data looking at the effect of land-use and landscape composition on the efficiency on interventions suggested to enhance biodiversity and ecosystem delivery in agricultural landscapes. This project will apply a Bayesian Evidence Synthesis approach to perform a cost-efficiency analysis. The approach is a well described methodology suitable when an analysis requires simulations informed by evidence. There is a growing interest of applying BES on this type of problems.  The student will learn Bayesian methods and perform the analysis in OpenBugs and R. JAGs is also a possibility.

Master project 4. Statistical testing without p-values using Bayesian Data Analysis

This project will critically evaluate hypothesis testing using a Bayesian approach. The student can choose a theory (model) to test and perform his/her own field experiment or get access to available empirical data for which a theory is to be tested. The main focus is the methodological work to compare classical statistical with Bayesian testing and perform the testing. The student will need learn Bayesian methods and perform the analysis in R.