Bio: Hi! I’m a second-year Maths and Stats undergraduate at St John’s College, University of Oxford. I’m very excited for my upcoming project with Dr Ioanna Manolopoulou on binary classification problem with datasets containing misdiagnoses – for me, this is a valuable experience in statistical research in the context of health science, and I look forward to honing my theoretical understanding and computational skills over the internship. Currently, I’m interested in mathematical and statistical methods and their applications, and I anticipate studying statistical models, mathematical biology and the qualitative theory of differential equations in depth over the next two years.
Project: Modelling disease with non-uniform (mis)diagnostic rate with Dr Ioanna Manolopoulou at UCL.
Modelling a binary outcome (such as the presence or absence of disease) given a set of explanatory factors is a well-studied problem in statistical literature. However, modelling and inference becomes much more challenging if the 0/1 labels are observed with error, especially if the error may also depend on the covariates of interest. For example, diagnostic devices might have higher accuracy for people with certain characteristics, or diagnostic decision trees might tend to identify patients with particular disease presentation. Depending on the extent of the mislabelling, ignoring the observation error could lead to vastly inaccurate inference results and models that do not reflect the true outcomes. This is important both because it leads to misdiagnoses, but also because it biases the data we use to improve models in the future.