Statistical models are valuable tools for informing conservation practices; their outputs often guide policy decisions, identify key driving mechanisms, and shape strategy. Consequently, the demand for accessible methods that produce interpretable results has increased substantially. This thesis addresses the demand from two perspectives. First, to enable identification of sensitive or illegal behaviours which affect biodiversity, we develop a comprehensive framework categorizing all Randomised Response Techniques (RRTs) where the sensitive variable is discrete. We demonstrate how this framework unifies several classical discrete RRT designs, enabling a straightforward comparison of their properties. Building upon this, we introduce a procedure that allows researchers to design surveys optimised for efficiency while meeting specified privacy constraints. To support practical implementation, this work is accompanied by an accessible RShiny application. Second, to aid in identifying common mechanisms of change, we develop a hierarchical clustering framework that aims to balance the minimisation of within-cluster variation with the interpretability of predefined ecological groupings. We apply this framework to a large-scale dataset on Canada's bird populations. Our analysis reveals that traditional guild-level summaries miss substantial species-level variation and proposes several alternative clustering strategies.
Thomas Cheale (Mon,) studied this question.