Model Selection and Model Simplification Model Selection and Simplification in R – a live online course covering model fit, nested model comparison, cross-validation, information criteria (AIC/BIC), and variable selection methods including stepwise, ridge, Lasso, and elastic net.
https://www.prstats.org/course/model-selection-and-model-simplification-msms05/ Introducing *Model Selection and Model Simplification (MSMS05)* — a live, hands-on two-day course that gives you the tools to make smarter, more reliable statistical models. ------------------------------ Why this course matters - In real-world data analysis, you rarely want the most complex model — you want the best model. This course teaches how to *determine which terms to keep, which to drop, and how to assess competing models* — all in a principled way. - Overfitting or underfitting can mislead decision-making or scientific inferences. You’ll learn when a more complex model actually harms predictive performance (and when it doesn’t). - The methods covered (AIC, BIC, cross-validation, nested model comparison, penalised regressions like ridge, Lasso, elastic net) are widely used in applied statistics, machine learning, ecology, biomedicine, economics, and beyond. - You’ll work in *R*, applying these techniques to realistic datasets. Code, datasets, and slides are all supplied, and you’re encouraged to bring your own data into the course. ------------------------------ What you will learn - Core metrics of model fit: likelihood, deviance, residual sums of squares - Techniques for *nested model comparison* (linear, GLMs, mixed-effect models) - Methods to evaluate *out-of-sample predictive performance* - Implementation and interpretation of *penalised regression methods* (ridge, Lasso, elastic net) - Principles and use of *model averaging* - Best practices: when to simplify versus when to keep complexity ------------------------------ Format & logistics - *Dates*: 3–4 November 2025 - *Duration*: 2 days, 4 hours each day - *Format*: Live online (all sessions recorded and a further 30 days access after the course - *Fee*: £250 - *Audience*: Data analysts, researchers, postgraduate students with some familiarity with R - *Support*: 30 days of post-course email support + access to recordings Email [email protected] with any questions. -- Oliver Hooker PhD. PR stats [[alternative HTML version deleted]] _______________________________________________ R-sig-ecology mailing list [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
