How large is large enough? Sample size calculations for clinical prediction models
Presenting a general framework for assurance-based sample size calculation across different modelling strategies.
Clinical prediction models are increasingly used to support decision-making, yet guidance on how large a dataset is needed to develop a reliable model remains limited. In practice, sample size is often not considered and is constrained by available data, with little consideration of the consequences for model performance and reliability.
In this talk, Dr Rebecca Whittle will discuss why sample size is a critical, yet frequently overlooked, aspect of prediction model development and evaluation. Insufficient sample sizes can lead to datasets that are not representative, unstable predictor effects, imprecise and miscalibrated predictions, and ultimately reduced clinical utility. She will briefly outline these key challenges before focusing on recent methodological developments.
Rebecca will present stability-based approaches to sample size determination, which aim to ensure that model predictions are sufficiently precise. She will then introduce a general framework for determining sample size across different modelling strategies, including both traditional regression-based statistical models and machine learning approaches.
Together, this work focuses on reducing instability in prediction models and moving beyond approaches that target performance on average, towards assurance-based sample size calculations that ensure models meet desired criteria with high probability.
Speaker
- Dr Rebecca Whittle, Medical Statistician, University of Birmingham
Event notices
- Please note that you can join this event in person or you can join the session remotely
- Please note that the recording link will be listed on this page when available
Admission
Contact
