----------------------------------------------------------------------------- A v v i s o d i M i n i - C o r s o ----------------------------------------------------------------------------- Dipartimento di Scienze Statistiche Sapienza Università di Roma ----------------------------------------------------------------------------- Mercoledì, 13 Dicembre, ore 08:30-12:30 (Aula VI, 4 piano) Giovedì, 14 Dicembre, ore 15:00-19:00 Venerdì, 15 Dicembre, ore 15:00-19:00 ----------------------------------------------------------------------------- Il corso è gratuito, prevalentemente rivolto agli studenti del curriculum internazionale della Laurea Magistrale in Statistics and Decision Sciences, ma aperto anche alla partecipazione di esterni. Dato il numero limitato di posti disponibili in aula, tutti gli interessati sono invitati a **prenotarsi** compilando il seguente form entro domenica 10 Dicembre: url: https://goo.gl/vfM9FM Entro lunedì 11 Dicembre verrà inviata conferma dell'effettiva disponibilità del posto all’indirizzo email fornito. Ovviamente “first come, first served”, con possibile piccolo bias in favore di studenti e dottorandi. Buon ponte lungo... Pierpaolo Brutti ----------------------------------------------------------------------------- W H O / Haakon Bakka (https://www.ntnu.edu/employees/haakon.bakka) T I T L E / Introduction to R-INLA for Bayesian Spatial Modelling A B S T R A C T / The course starts with a general overview of the possibilities in INLA, for applied research and for model development. We will proceed to examples of generalised linear models with several random effects. We discuss the cool ideas making INLA fast; why most likelihoods are near-Gaussian in the posterior, how to represent random effects with sparse matrices, and more. The last part of the course is looking at spatial random effects and how they fit into the INLA framework; using the brilliant SPDE approach. In the SPDE approach we approximate the spatial random effect itself, representing it in the entire study area, without needing to know the observation locations. Inference in INLA is so fast that we can run all examples live in class! I am looking forwards to giving the course, and I hope to see you there. S Y L L A B U S / Introduction - What is INLA? Why are so many using it? Why the high impact? - Why you should be very excited to be in this course! - The two core ideas: GMRF, Laplace - Bayesian fitting and prediction Time series and several nonlinear effects - Data: Unemployment - INLA inputs and outputs - GMRF structure with spare precision matrix - Run time O(N) - Simulation-inference Exploring hyper-space - Conditionally Gaussian - Gradient descent - Good computational parametrisations - Logfile & Inference diagnostics - Representing the posterior NOT as samples but as... Non-Gaussian likelihoods - Data: Seeds - Laplace approximation / Nested Laplace - Simulation-inference What models exist in INLA - What likelihoods and links? - What happens when we change the link function? - Multiple likelihoods - Rgeneric: Every model is possible! Priors and interpretable parameters - Interpretable parametrisation vs computational parametrisation - Uninformative priors, conjugate priors - Good priors! - Hierarchical models and “analysis of component variance” Disease-mapping with Besag - Spatial model - Revisiting all the topics so far Continuous spatial models - Data: Fish larvae - How to get O(N^1.5) instead of O(N^3) - De-connecting observation locations and the underlying model - The famous SPDE-approach - Finite Element method and mesh - Simulation-inference -----------------------------------------------------------------------------