A PSI Training Course in
The Analysis of Continuous and Discrete Longitudinal Data
GlaxoSmithKline Professor of Biostatistics
London School of Hygiene and Tropical Medicine
20 / 21 March 2012
This course provides an introduction to the analysis of longitudinal and repeated measurements data, principally in a clinical trial setting.
Day 1 – Continuous Outcomes
The first day will cover the analysis of continuous outcomes. The well-known framework of the multivariate normal linear model will be used, and the main modelling and interpretational issues will be covered. The SAS procedures MIXED and GLIMMIX will be used to illustrate the examples. Key points covered will be:-
·Extending the linear model: random effects, random coefficients, and mixed effects models;
·Covariance structure: its role in the analysis, and choice of structure;
·Appropriate inference, small sample issues, and sample size calculations.
Day 2 – Discrete Data
On day two, consideration will be given to discrete data, focusing principally on binary outcomes. Much of this development applies equally well to ordinal and count outcomes, additional considerations only for these settings will be highlighted. Such analyses raise several important issues that are not a concern with the analogous analyses based on linear models. There is no single natural modelling framework for such data and different routes to analysis involve models that have quite different interpretations. The most important distinction is that between so-called population averaged (or marginal) and subject-specific (or conditional) models. A careful introduction to the whole area will be given. Applications will again be explored and illustrated using several SAS procedures (GLIMMIX, GENMOD, NLMIXED), placing emphasis on the comparative interpretation of the different classes of model. The main points covered will be
·Issues with dependent discrete data, with an introduction to, and comparison of, the main classes of model;
·Marginal models for binary and binomial data;
·The empirical estimate of error;
·Subject-specific models for binary and binomial data;
·Issues specific to ordinal and count outcomes.
Throughout the focus will be on data from longitudinal and cross-over clinical trials, and the choice of illustrative examples will reflect this. It will be assumed that participants have experience with conventional linear regression and generalised linear models.
About the presenter:
Mike Kenward has been the GSK Professor of Biostatistics at LSHTM since 1999. He has over 25 years experience of consulting within the pharmaceutical industry, and in that period has given over 100 short courses worldwide on a variety of topics, including several for PSI. He is co-author of two standard texts: The Design and Analysis of Cross-Over Trials (with Byron Jones) and Missing Data in Clinical Studies (with Geert Molenberghs). He is currently writing a book on Multiple Imputation with James Carpenter.
Course runs from:10:00 – 17:00 (registration from 09:30) on Day 1 and 09:00 – 16:30 on Day 2.
Please register online at www.psiweb.org and click on PSI Events; payment available online.
Registration costs (includes lunch and refreshments)
PSI Members: £595 plus vat
Non-members: £630 plus vat
PSI aims to be fully inclusive and endeavours to accommodate delegates with disabilities wherever possible. Please help us to help you by letting us know if you require additional facilities or have any special requirements. Please contact us on +44 (0)845 1800 349 or at [log in to unmask] for further information.
Tel: +44 (0)845 1800 349,
Fax: +44 (0)1730 715291
Email: [log in to unmask]
Statisticians in the Pharmaceutical Industry Limited
Registered Office: Durford Mill, Petersfield, Hampshire GU31 5AZ
A Limited Company registered in England and Wales No. 4634899
You may leave the list at any time by sending the command
to [log in to unmask], leaving the subject line blank.