Funded PhD Studentship in Statistics and Data Science

Project description

Complex systems are central to understanding many real-world challenges. Real-world data sets, such as road accidents or property transactions, often exhibit segmented behaviour where variance changes near critical points. These systems often exhibit fluctuating behaviours and critical transitions, with variance and skewness playing a critical role in their analysis.

Linear regression models often fail to account for key features like heteroscedasticity (variance that depends on the signal) and skewness, which can lead to biased model parameters. These issues also arise in other complex data analysis problems across areas such as public health, finance, and environmental science. For example, floods, which are becoming increasingly common due to climate change, and wind speeds, critical with the growing reliance on renewable energy, both require models with flexible distributions to account for their varying variance and asymmetry.

By improving these models, we can enhance our ability to predict future events and allocate resources more effectively to mitigate their impacts and protect communities. 

Purpose/objectives 

1. Segmented Time-Dependent Data Modelling: Develop models that allow for segmented relationship while accounting for skewness and heteroscedasticity over time. 
2. Expanding Distributional Flexibility: Enhance the Bayesian framework by incorporating a broader range of analytically intractable distributions. This will improve the framework’s ability to accurately represent complex datasets related to flooding and wind speed, thereby providing greater flexibility and accuracy in predictions. 
3. Develop Predictive Models: Use the expanded Bayesian framework and machine learning methods to develop predictive models for flood risk and wind dynamics. 

Principal accountabilities and responsibilities

In this PhD project, you will develop and apply an advanced Bayesian Generalised Modelling Framework (BGMF) that allows for segmentation and time dependency accounting for variance and skew. Additionally, you will expand the BGMF to include a wider range of analytically intractable distributions. This enhancement will improve the model’s ability to represent complex datasets related to flooding and wind speed, thereby offering greater flexibility and accuracy in predictions.

Finally, apply the enhanced BGMF along with machine learning methods, to real-world data provided by the Met Office (e.g., temperature, rainfall, sunshine, radiation, wind) and the Environment Agency (e.g., water level, flow, wind, temperature), with the aim of improving forecasting accuracy and optimising resource allocation to mitigate the impacts of these natural phenomena.

To apply

For more information and to apply, please visit our website: https://www.derby.ac.uk/research/degrees/applicants/studentship-opportunities/mphil-phd-studentship-statistics-and-data-science/

Closing dates for applications: 5pm, Monday 6 January 2025.

Interviews: 22 January to 23 January 2025.

For informal enquires please contact Dr Jack Sutton, Lecturer in Statistics and Data Science via: [Email Address Removed]