The US National Trauma Database is the major repository including 2 million records of trauma patients, available for the research community. To predict patient's outcomes practitioners use Decision tree (DT) models which can be efficiently built on given data and provide a transparent interpretation within a probabilistic framework. Averaging over DT models under certain conditions can deliver reliable information on predictive posterior probability distributions, which is of critical importance in the case of predicting a patient's outcome. Reliable estimations of the distribution can be achieved within the Bayesian framework using Markov chain Monte Carlo (MCMC) and its Reversible Jump extension enabling DT models to grow to a reasonable size. Existing MCMC strategies however have limited ability to control DT structures and tend to sample overgrown DT models making unreasonably small partitions, thus deteriorating the uncertainty calibration. This happens because the MCMC explores a DT model parameter space within a limited knowledge of the distribution of data partitions. The project aims to explore new strategies which can be adapted to variable data distribution in order to efficiently overcome the existing limitation. It is expected that in the case of predicting trauma outcomes the number of data partitions can be significantly reduced. This will reduce the unnecessary uncertainty of estimating the predictive posterior probability density.
Research Questions: (1) to explore the ability of designed strategies to extend the prediction of trauma survival (2) to explore ways of designing reliable decision models within the Bayesian framework.
The deadlines are as follows:
For March starters:
International applicants - 30th November 2021
UK nationals - 18th January 2022
For October starters:
International applicants - 30th June 2022
UK nationals - 5th August 2022