Workshop on Vector Generalized Linear and Additive Models
14–16 June 2016, University of Cagliari
by T. W. Yee (University of Auckland)
Thomas W. Yee, Senior Lecturer at Auckland University, is currently Visiting Scientist at the University of Cagliari in the framework of the Visiting Scientist Program supported by the Regione Autonoma della Sardegna.
He will present a 3-day workshop intended as an applied course on regression modelling, aimed towards young researchers, PhD students and postgraduate students in statistics. The material is selected from the book “Vector Generalized Linear and Additive Models: With an Implementation in R” recently published by Springer, and involves class lectures and computer tutorial labs—for a total of 12 hours (4 hours per day).
A few topics will be dealt with more theoretically, however most of the course is aimed to be practical. The morning sessions will involve presentation of the ‘theory’ side and problem solving through the analysis of data sets.
Afternoon sessions might be organized on request to discuss about theoretical issues presented in morning sessions and/or analyze real problems via the VGAM R package.
1st day: Revision and Introduction to Vector Generalized Linear Models (VGLMs)
2nd day: Reduced Rank-VGLMs, and Applications to Count Regression
3rd day: Further topics and Applications (Varying Coefficient Models, Categorical Regression Models, Extremes,Univariate and Distributions)
Dipartimento di Scienze Economiche e Aziendali, University of Cagliari
Computer Room, Viale Frà Ignazio 74, 09123, Cagliari
14 June: 9 am – 1 pm (+ 3 pm – 5 pm, on request)
15 June: 9 am – 1 pm (+ 3 pm – 5 pm, on request)
16 June: 9 am – 1 pm (+ 3 pm – 5 pm, on request)
The participation to the workshop is free (no registration fee). For organizational reasons interested students should confirm their participation by email to email@example.com (subject: VGLM workshop).
Day 1: Revision and Introduction to VGLMs
1.1 Revision of parts of R, LMs, GLMs (e.g., formulas, lm(), glm())
1.2 Revision of Smoothing (regression splines, bs(), ns())
1.3 Six Illustrative Models (e.g., uninormal(), negbinomial(), multinomial())
1.4 Introduction to VGLMs and VGAMs (definitions, simple examples)
1.5 More on VGLMs and VGAMs ( Constraint Matrices, xij, IRLS, EIMs)
1.7 Using the VGAM R package
Day 2: RR-VGLMs, and Applications to Count Regression
2.1 RR-VGLM Definitions
2.2 Examples (including Row-Column Interaction Models)
2.3 Counts: Negative Binomial Regression
2.4 Positive Count Distributions
2.5 Zero-inflated and Zero-altered Models
2.6 Writing VGAM Family Functions (e.g., EIMs and S4)
Day 3: Some Further Applications
3.1 Varying Coefficient Models (normal.vcm())
3.2 Categorical Regression Models (e.g., acat(), cratio(), cumulative())
3.3 Extremes (gev(), gpd())
3.4 Univariate Distributions
3.5 Bivariate Distributions (e.g., copulas)
Background and Preparation
Students are assumed to have a basic working knowledge of R, as well as some experience fitting simple linear models and generalized linear models (logistic regression and Poisson regression). A good book for background reading in this area is Fox and Weisberg (2011). (The books Faraway (2015) and Faraway (2006) may also be useful.)
Why R? R is freely available from www.R-project.org and runs on all popular operating systems such as Windows, Macintosh and Linux. Also it is open-source, fully-featured (including 7000+packages), and is very powerful and has superb graphics. Most academic statisticians worldwide use R, and it is usually the first platform where new statistical methodology is implemented.
Participants are encouraged to bring their own laptops with the latest version of R installed. The university computer laboratories should be available for students without a laptop. Also, the latest version of the following R packages should also be installed: VGAM, VGAMdata. The following packages may also be useful: COUNT, AER, mlogit.