Course overview
Predicting the unobserved has become an increasingly popular tool in statistics. It can provide a unifying framework to achieve many goals in statistical analysis - imputing missing data, predicting the future, predicting unobserved members of the population and what the impact of a particular treatment would be if it was administered. In this course students will build upon a core understanding of these specialist statistical areas to unify them with a central theory. In doing so, students will embed cutting edge research in assessing the quality of prediction and contemplate current challenges in the field.
Course learning outcomes
- Distil common statistical tasks into a predictive task using Ruben’s potential outcomes framework
- Analyse data with structure for a specific prediction task using Bayesian inference or machine learning
- Propose validation and scoring schemes for different structures of data and prediction tasks
- Interpret original research articles in the context of an applied problem
- Translate approximate cross-validation algorithms into code
- Present predictions to different audiences with a particular focus on limitations and assumptions
Degree list
The following degrees include this course