20-24 November 2023
To foster international participation, this course will be held online
Most of the statistical methods you are likely to have encountered will have specified fixed functional forms for the relationships between covariates and the response, either implicitly or explicitly. These might be linear effects or involve polynomials, such as x + x2 + x3. Generalised additive models (GAMs) are different; they build upon the generalised linear model by allowing the shapes of the relationships between response and covariates to be learned from the data using splines. Modern GAMs, it turns out, are a very general framework for data analysis, encompassing many models as special cases, including GLMs and GLMMs, and the variety of types of splines available to users allows GAMs to be used in a surprisingly large number of situations. In this course we’ll show you how to leverage the power and flexibility of splines to go beyond parametric modelling techniques like GLMs.
The course is aimed at graduate students and researchers with some statistical knowledge; ideally you'll know something about generalized linear models, likelihood, and AIC. However we’ll recap
what GLMs are so if you’re a little rusty or not everything mentioned in the GLM course makes sense, we have you covered. From running the course previously, knowing the difference between
"fixed" and "random" effects, and what the terms "random intercepts" and "random slopes" are, will be helpful for the Hierarchical GAM topic, but we don't expect you to be an expert in mixed
effects or hierarchical models to
take this course.
Participants should be familiar with RStudio and have some fluency in programming R code, including being able to import, manipulate (e.g. modify variables) and visualise data. There will be a mix of lectures, in-class discussion, and hands-on practical exercises along the course.
1. Understand how GAMs work from a practical view point to learn relationships between covariates and response from the data
2. Be able to fit GAMs in R using the mgcv and brms packages
3. Know the differences between the types of splines and when to use them in your models
4. Know how to visualise fitted GAMs and to check the assumptions of the model
Sessions from 14:00 to 20:00 (Monday to Thursday), 14:00 to 19:00 on Friday (Berlin time). From Tuesday to Friday, the first hour will be dedicated to Q&A and working through practical exercises or students’ own analyses over Slack and Zoom. Sessions will interweave mix lectures, in-class discussion/ Q&A, and practical exercises.
Monday– Classes from 2-8 PM Berlin time
Brief overview of R and the Tidyverse packages we’ll encounter throughout the course
Recap generalised linear models
Fitting your first GAM
Tuesday– Classes from 2-8 PM Berlin time
How do GAMs work? What are splines? How do GAMs learn from data without overfitting? We’ll dig under the hood a bit to understand how GAMs work at a practical level and how to use the mgcv and gratia packages to estimate GAMs and visualise them.
Wednesday– Classes from 2-8 PM Berlin time
Model checking, selection, and visualisation. How do we do inference with GAMs? Go beyond simple GAMs to include smooth interactions and models with multiples smooths.
Thursday– Classes from 2-8 PM Berlin time
Hierarchical GAMs; introducing random smooths and how to model data with both group and individual smooth effects.
Doing more with your models; introducing posterior simulation
Friday– Classes from 2-7 PM Berlin time
Fitting GAMs with brms
Going beyond the mean; fitting distributional models and quantile GAMs
Should you have any further questions, please send an email to firstname.lastname@example.org
> 30 days before the start date = 30% cancellation fee
< 30 days before the start date= No Refund.
Physalia-courses cannot be held responsible for any travel fees, accommodation or other expenses incurred to you as a result of the cancellation.