Clinical Data Science / Course details

Clinical Data Scientists are required to generate insights from data, create or contribute to data products and interpret/translate clinical research. In order to achieve this they require a working knowledge and experience of data analysis methods including statistical learning (statistics and machine learning methods) supported by knowledge and understanding of the mathematical principles underpinning these methods and in which situations to apply them, including how to select the most appropriate method(s) based on the nature of the problem, data available and clinical/organisational priorities. 

The unit aims to:

• Provide the underpinning applied mathematical concepts to common data science and machine learning methods
• Provide students experience and practice with running analysis scripts (using R and Python) to analyse datasets using statistical and machine learning methods and interpreting output
•  Expose students to a variety of analytic methods for tackling a range real world clinical problems

This unit will cover the following indicative content:

• Fundamental applied mathematics for data science
• Examples of using R and Python for data science
• Introduction to inferential statistics and hypothesis testing
• Introduction to machine learning algorithms for classification and regression problems
• Selecting the appropriate analysis method and evaluating model performance
• Issues around bias in data and wider ethical and legal issues related to automation of tasks with machine learning and their impact on patient outcomes

The unit will be delivered online making use of workshops, lectures, self-directed learning material delivered through interactive digital (Jupyter) notebooks and synchronous labs helping students to work through analysis of various data sets.

LO1: Explain the underpinning mathematical concepts behind commonly used statistical and machine learning methods (e.g. probability theory, linear algebra)

LO2: Critique different approaches to analysis based on the data available and clinical/organisational goals

LO3: Identify the main differences, requirements and caveats in analytical approaches depending on intended goals and properties of the data (e.g. hypothesis testing vs data-driven approaches) 

LO4: State the stages involved in a typical machine learning analysis pipeline

LO5: Interpret and annotate model output providing context to convey findings to a variety stakeholders

LO6: Evaluate different statistical and machine learning/NLP approaches and select appropriate methods depending on the properties of the available data and task requirements

LO7: Evaluate the outputs of statistical tests and machine learning models and report these results using standard metrics

LO8: Work through the problem-solving cycle

LO9: Develop an analytical problem solving mind-set

Formative assessment and feedback to students is a key feature of the on-line learning materials for this unit and is provided through self-directed learning activities in the interactive notebooks. 

• Field, A., Miles, J., Field, Z (2012) Discovering Statistics Using R. Los Angeles: SAGE
• James, G., Witten, D., Hastie, T., Tibshirani, R (2015) An Introduction to Statistical Learning: With applications in R. New York: Springer
• Géron, A (2017) Hands-on Machine Learning with Scikit-Learn & TensorFlow. Beijing: O'Reilly
• Lane, H., Howard, C., Hapke, H (2019) Natural Language Processing in action. Shelter Island: Manning Publications Co.