General & Conceptual

Question 1

What motivated you to pursue this topic?

Answer 1

• compositional data is an interesting topic found in a range of real-world applications
• typical addressed through modelling approach (log-ratio) and most of the research has followed this
• traditional approaches to compositional data are often impractical - zeros, missing, counts
• flexible approaches to overcome these limitations, enabling richer and more interpretable analysis across a range of domains.

Question 2

What are the main research questions you addressed?

Answer 2

• modelling compositional data which contains a features that prevents the current proposed approach and contains a data challenge that further complicates the modelling problem

Question 3

Can you summarise your thesis in one sentence?

Answer 3

• applies novel Bayesian hierarchical frameworks for analysing compositional data with specific features and data challenges which prohibit log-ratio transformations from being applied

Question 4

What are the main limitations of your work?

Answer 4

• computational cost - impact scalability for high-dimensional data
• cannot share data for full reproducibility