Thesis Abstract
The increased availability of clinical data, in particular case data collected routinely, provides a valuable opportunity for analysis with a view to support evidence based decision making. In order to confidently leverage this data in support of decision making, it is essential to analyse it with rigour by employing the most appropriate statistical method. It can be difficult for a clinician to choose the appropriate statistical method and indeed the choice is not always straight forward, even for a statistician. The considerations as to what model to use depend on the research question, data and at times background information from the clinician, and will vary from model to model.
My thesis develops an intelligent decision support method that supports the clinician by recommending the most appropriate statistical model approach given the research question and the available data.
The main contributions of my thesis are:- Identification of the requirements from real-world collaboration with clinicians.
- Development of an argumentation based approach to recommend statistical models based on a research question and data features; an argumentation scheme for proposing possible models.
- A statistical knowledge base designed to support the argumentation scheme, critical questions and preferences.
- A method of reasoning with the generated arguments and preference arguments.
- The approach is evaluated through case studies and a prototype.
