Prediction Model Performance for Binary Outcomes in R
Interested in learning how to evaluate the performance and reliability of prediction models? Assessing model performance is a critical step in determining whether a prediction model is accurate, trustworthy, and suitable for research or clinical applications. Through live demonstrations and hands-on exercises in R, participants will gain practical experience using widely accepted methods to evaluate binary outcome prediction models. This workshop introduces key concepts and techniques for assessing prediction model performance, including discrimination, calibration, and internal validation. Participants will learn how to generate and interpret common performance metrics and create clear, publication-ready visualizations to communicate their findings effectively.
By the end of this workshop, participants will be able to:
- Develop prediction models for binary outcomes and generate predicted probabilities in R.
- Assess model discrimination using the C-statistic (AUC) and receiver operating characteristic (ROC) curves.
- Evaluate model calibration using calibration plots and calibration statistics.
- Perform internal validation using resampling techniques such as bootstrapping and cross-validation.
- Create publication-ready figures, tables, and summaries to effectively communicate prediction model performance.
This session is ideal for researchers, graduate students, and analysts interested in developing and evaluating prediction models. Basic familiarity with R and regression modeling concepts is recommended.
Workshop Preparation
None
Facilitator Bio
Shubrandu (he/him) is a PhD Candidate in Health Research Methodology (Clinical Epidemiology) at McMaster University with a background in epidemiology and biostatistics. He has extensive experience working with large healthcare datasets and supports researchers in study design, statistical analysis, and data management. His expertise includes regression modeling, survival analysis, and handling administrative health data, with proficiency in R and SAS.
Workshop Slides
Coming soon.