University of Edinburgh
School of Mathematics and BioSS
Date: Friday 5th May, 15:10 Location: JCMB 5323
Speakers: Professor Jane Hutton, University of Warwick and Professor Anthony Davison, EPFL, Switzerland
Professor Jane Hutton
Title: Missing data and how to see biased results
Abstract:
Any study of interventions which take time to have an effect, whether treatment for cancer or diets for weight loss, require measurements, often several measurements to be recorded at some interval after the initial assessment. For some conditions, such as diabetes, it is reasonable to expect people to visit a clinic regularly, but for minor conditions it is usual to rely on mail or telephone questionnaires. Study participants might be too busy to complete the questionnaire, might think their reply is not important, might have moved, or might have died. What are the possible biases in our conclusions if we summarise only the available data? If one treatment leads to quicker recovery, and those who have recovered are less likely to respond, then it is very difficult to obtain a reliable estimate of treatment effect.
This talk will begin with common definitions of missing data, which reflect the situations described above. Chain Event Graphs will then be introduced: they are statistical models for a set of random variables whose joint probability function is described in terms of a graph. The graphs are derived from probability trees by merging nodes in tree with the same associated conditional probabilities. This results in an accessible, visual representation of the statistical model.
Professor Anthony Davison
Title: Statistical models for complex extreme events
Abstract:
Awareness of the importance of risk assessment for complex extreme events has greatly increased in recent times. Leading examples of such events are floods and heatwaves, attributed to global change, and crashes in financial markets, often attributed to mis-estimation or mis-understanding of the dependence between financial actors such as banks. Probability models for univariate events are well-established, and those for multivariate events are in principle well-understood, but useful statistical models and efficient methods of inference in more complex settings are less well developed. This talk will give an overview of some recent work in the area, with an emphasis on environmental applications.
This seminar is a part of Maxwell Institute seminar series.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|