Skip to main content

Advanced Modelling Strategies: challenges and pitfalls in robust causal inference with observational data

Training Event
Monday 17 - Thursday 20 July, 2017, 9am - 5pm
Leeds Institute for Data Analytics, Level 11, Worsley Building, University of Leeds, Clarendon Way, Leeds, LS2 9NL

**Please note that this course is now full**

This four day summer school, supported by SSM, is designed to give an introduction to the common pitfalls and challenges in statistical multivariable regression modelling of observational data.  The school is run by Prof Mark S Gilthorpe (Medicine, Leeds) with input from Dr Peter WG Tennant (Healthcare, Leeds) and Dr George TH Ellison (Medicine, Leeds) and is based on materials prepared in conjunction with Dr Johannes Textor (Radboud University Medical Center, Nijmegen).

By the end of the course, participants will be able to critically appraise a wide range of applications of statistical regression models as presented in the literature with respect to the the exploration of complex, potentially causal relationships, including the longitudinal analyses of change, mediation, non-linearity and statistical interaction.

Knowledge will be developed that helps identify modelling strategies that are potentially erroneous and understanding alternative strategies (if they exist) that avoid the adverse impacts of mathematical coupling, the family of paradoxes called the reversal paradox (that include Simpson's paradox, Lord's paradox and suppression), compositional data, and inappropriate employment of ratio variables.  This will all be set within a framework of seeking robust causal inference, with a blend of lectures and workshops that provide first-hand insight to practical worked examples.

Many of the workshop examples that are presented in this course will be taken from the epidemiological literature, though the same multivariable statistical modelling strategies engaged will be familiar to a variety of other disciplines, and hence the relevance of the methodology and application within the course is not restricted to health researchers. Workshops use the statistical package R, though all R code is provided as focus is not the statistical programming but understanding the concepts that are demonstrated by execution of the R code provided. Please get in touch if you have any questions about the suitability of the course.


  • promote a sound rationale with respect to modelling causal inference of observational data;
  • introduce common pitfalls and challenges in the statistical modelling of observational data;
  • develop an understanding of underlying assumptions to various modelling strategies;
  • develop the knowledge-base of the various pitfalls in statistical modelling;
  • understand situations in which certain modelling strategies are erroneous or misleading;
  • evaluate new modelling strategies based on an inherently questioning stance to methodology;
  • know existing or currently emerging statistical methods relevant to longitudinal data analysis;
  • appreciate the importance of data generation in the selection of appropriate statistical models.

Transferable skills

By the end of this course the participants should:

  • know how to set about statistical modelling of observational data with respect to causal inference;
  • know what methodologies not to employ in a range of frequently encountered circumstances;
  • comprehend the complex analyses provided by multivariable modelling of observational data;
  • be able to undertake the critical appraisal of modelling strategies of other researchers.


Student rate: £225
Other: £450

SSM Funding

Funding was available from SSM for 6 ECR & 6 MCR researchers, please note that only 1 funded MCR space remains.  Please indicate on the form if you wish to apply for the funded places, these will be allocated on a first come first serve basis; you must have been an SSM member since the beginning of this year.


Please note that this course is now full. For enquiries about the course, please contact Eleri Pound.