An Introduction to Multilevel
Modelling

The Multilevel Modelling course is now full and we regret that we can no longer accept registrations for the course.

For participants registered in the course please follow this link to a page that provides information on how to prepare for the course.

Course Description

Data that are aggregated at multiple levels are common in many aspects of health research, particularly in veterinary medicine in which animals may be grouped within pens which are in turn grouped within herds which are grouped by geographic region. Unfortunately, standard multivariable modelling techniques such as linear and logistic regression which ignore this natural clustering, may provide incorrect estimates of parameters and their standard errors. However, over the past 5 years, there has been a tremendous growth in the availability of multilevel modelling techniques to appropriately handle these types of data.

Through a series of lectures and laboratory sessions, this course will provide students with a solid understanding of how to develop and interpret multilevel models. Topics to be covered will include:

-Introduction to correlated data
-Introduction to generalized linear models (GLM)
-Fitting linear multilevel models (for continuous data)
-Evaluating sources of variation in multilevel models
-Fitting discrete multilevel models (binomial data)
-Evaluation of models (diagnostics)
-Advanced procedures for fitting discrete models (bootstrapping and simulation methods)
-Alternative approaches to dealing with clustered data
-Dealing with the special case of repeated measures

Most of the laboratory sessions will utilize the commercially available software MLwiN, a very flexible, yet intuitive program for building multilevel models. Complementary approaches for analysing these types of data in either Stata or SAS will also be presented.

Students are strongly encouraged to bring their own data to the course. Assistance will be provided in formatting these data to allow for their analysis using MLwiN. Throughout the course, students will be provided with an opportunity to work on their own data (or on examples provided if they do not have any data of their own to analyse) and the last day of the course will be devoted to a discussion of a number of these analyses.

Who Should Attend

This course will be oriented to anyone who routinely

analyses health related data in either a research or control program setting,

or

need to understand and interpret the results of such analyses as part of their regular duties.

Students will be expected to have a working knowledge of ordinary linear and logistic regression. Some background reading material will be provided to all participants, prior to the course.

Instructors

Dr. Henrik Stryhn, MSc, PhD Danish Veterinary Laboratory, Copenhagen, Denmark Dr. Stryhn is a biostatistician who currently holds a position as a senior research scientist in the Department of Pathology and Epidemiology in the Danish Veterinary Laboratory. He has conducted research and published in the area of multilevel models for discrete data. He has also distinguished himself as an effective teacher in a number of short courses covering this material. More information on Dr. Stryhn’s background can be found at: http://www.dina.dk/~hes/
Dr. Ian Dohoo, DVM, PhD Atlantic Veterinary College, University of PEI Charlottetown, PEI, Canada Dr. Dohoo is a Professor of epidemiology at the Atlantic Veterinary College. He has a special interest in analytic methods used in epidemiologic research and has a published a number of papers in this area (including multilevel models). He has participated extensively in the training of veterinary epidemiologists around the world.
	Drs Dohoo and Stryhn trying to find their way down the mountain
James Valcour, BSc. MSc Atlantic Veterinary College, University of PEI Charlottetown, PEI, Canada James Valcour is a recent graduate with an MSc in epidemiology. He currently provides technical support for epidemiologic research in the Dept. of Health Management at the Atlantic Veterinary College and he will provide technical assistance throughout the course.

Tentative Schedule (Subject to Revision)

(each block represents a 1.5 - 2 hour time slot)

For more information contact Ian Dohoo <dohoo@upei.ca> or James Valcour <jvalcour@upei.ca>