www.highstat.com
Highland Statistics Ltd

Online course: Introduction to regression models with spatial correlation using R-INLA

This online course consists of 9 modules representing a total of approximately 50-60 hours of work. Each module consists of video files with short theory presentations, followed by exercises using real data sets, and video files discussing the solutions. All video files are on-demand and can be watched online, as often as you want, at any time of the day, within a 12 month period.

A discussion board allows for daily interaction between instructors and participants. The course also contains a series of 2-hour web seminars in which we discuss some of the exercises. Attending these web seminars is optional. Depending on requests, we may run the web seminars multiple times per day and in different time zones. The web seminars are scheduled over a period of 3 weeks and will be repeated throughout the year.

You are invited to apply the statistical techniques discussed during the course on your own data and if you encounter any problems, you can ask questions on the Discussion Board.

A detailed outline of the course is provided below.

 

Part I: Limitations of frequentist approaches

Module 1: Introduction and revision.

  • A video with a general introduction and a discussion of dependency.
  • Revision exercise for multiple linear regression.
  • A short video explaining basic matrix algebra.

Module 2: Adding correlation to a regression model using frequentist techniques.

  • A video showing how to add temporal or spatial correlation to a linear regression model in a frequentist setting.
  • A video revising simple linear mixed-effects models.

Web seminar 1: A 2-hour web-seminar will be scheduled. In this web seminar, we will summarise modules 1 and 2.

 

Part II: INLA and the application of linear (mixed-effects) models in INLA

Module 3: Introduction Bayesian statistics.

  • A video presentation with an introduction to Bayesian statistics and the role of priors.

Module 4: Introduction to INLA.

  • A video presentation explaining the basic principles of INLA.

Web seminar 2: A 2-hour web-seminar will be scheduled. In this web seminar, we will summarise modules 3 and 4.

Module 5: Linear (mixed-effects) models with spatial correlation.

  • One exercise showing how to execute a linear regression model in INLA.
  • One exercise showing how to execute a linear mixed-effects model in INLA.

Web seminar 3: A 2-hour web-seminar will be scheduled. In this web seminar, we will discuss the exercises from module 5.

 

Part III: The application of GLM(M) in INLA

Module 6: Applications of GLM and GLMM in INLA.

  • Exercise showing how to execute a Poisson and negative binomial GLM in INLA.
  • Exercise showing how to execute a Bernoulli GLM in INLA.
  • Exercise showing how to execute a Poisson and Negative binomial GLMM in INLA.
  • Exercise showing how to execute a Bernoulli GLMM in INLA.

Web seminar 4: A 2-hour web seminar will be scheduled. In this web seminar, we will discuss the exercises from module 6.

 

Part IV: Spatial correlation in INLA and the application of GLM(M) with spatial correlation in INLA

Module 7: Applications of GLM and GLMMs with spatial correlation in INLA.

  • Exercise showing how to execute a Poisson and negative binomial GLM/GLMM with spatial correlation in INLA.
  • Exercise showing how to execute a Bernoulli GLM/GLMM in INLA with spatial correlation.
  • Exercise showing how to apply a beta GLM with spatial correlation in R-INLA.

Module 8: Dealing with natural barriers (e.g. an island for fisheries data).

  • Theory presentation of the barrier model.
  • Exercise showing how to deal with spatial correlation around an island for a coral reef data set.

Web seminar 5: A 2-hour web seminar will be scheduled. In this web seminar, we will discuss the exercises from modules 7 and 8.

 

 Part V: Regression models with spatial correlation for areal data

Module 9: Regression models and GLMs with spatial correlation for areal data.

  • Video file with theory presentation on the analysis of areal data.
  • Exercise showing how to execute a linear regression model with spatial correlation using areal data in INLA.
  • Exercise showing how to execute a GLM with spatial correlation using areal data in INLA.

 

Pre-required knowledge: Good knowledge of R, data exploration, linear regression and GLM (Poisson, negative binomial, Bernoulli). Working knowledge of mixed-effects models. Short revisions are provided. This is a non-technical course.

Cancellation policy: What if you are not able to participate? Once participants are given access to course exercises with R solution codes, pdf files of certain book chapters, pdf files of powerpoint files and video solution files, all course fees are non-refundable and non-transferable to another participant.

Copyright: Sharing the access details of the course website or the pdf files of our course material is prohibited. Video files cannot be downloaded, but they can be watched in the same way as on Netflix.

Web seminars: Web seminars are hosted on zoom.us. We will run each of them multiple times (3 - 4) per year, so you can choose when to attend (within the 12 months period). You can attend a specific seminar only once. In the first instance, we will limit the number of participants for a web seminar to 10. Seats are allocated on a first-come-first-serve basis. Please inform us by email when you would like to attend a specific seminar. A decent internet connection is required; see here for details (see the text under 'Recommended bandwidth for Webinar Attendees'). No refund will be given if your internet speed is too low.

Bring your own data: When you analyze your own data, it may be useful to have a little bit of help from us. You can use the Discussion Board for this. Note that this is not meant to be a statistical consultancy service, but rather a 'look-over-your-shoulder' type of help. To make this workable for us, we need a few common-sense rules. For example, we will not analyze your data, write R code for you, or execute your R code on our computer. Furthermore, your questions need to pertain to the course content (e.g. we will not answer questions on topics like zero inflation, GAMs, spatial-temporal correlation) and should be based on your own data; not that of your colleagues. If in our opinion, your questions diverge too much from the 'look-over-your-shoulder' type of help, we reserve the right not to engage any further with your data analysis. No refund will be given in such scenarios.