For further information on any CSM events please e-mail:

Due to impending industrial action the following CSM Seminar has been postponed. A new date will be arranged and announced in due course. Apologies for any inconvenience this may cause.

Centre for Statistical Methodology Seminar
Causal Inference Theme
Thursday 22 February 2018, 12:45-2:00pm – POSTPONED UNTIL A LATER DATE (TBC)
LG9, Keppel Street
How to obtain valid tests and confidence intervals for treatment effects after confounder selection?
Prof Stijn Vansteelandt (University of Ghent & LSHTM)

Abstract: The problem of how to best select variables for confounding adjustment forms one of key challenges in the evaluation of exposure or treatment effects in observational studies. Routine practice is often based on stepwise selection procedures that use hypothesis testing, change-in-estimate assessments or the lasso, which have all been criticised  for – amongst other things – not giving sufficient priority to the selection of confounders. This has prompted vigorous recent activity in developing procedures that prioritise the selection of confounders, while preventing the selection of so-called instrumental variables that are associated with exposure, but not outcome (after adjustment for the exposure). A major drawback of all these procedures is that there is no finite sample size at which they are guaranteed to deliver treatment effect estimators and associated confidence intervals with adequate performance. This is the result of the estimator jumping back and forth between different selected models, and standard confidence intervals ignoring the resulting model selection uncertainty. In this talk, I will develop insight into this by evaluating the finite-sample distribution of the exposure effect estimator in linear regression, under a number of the aforementioned confounder selection procedures. I will then make a simple but generic proposal for generalised linear models, which overcomes this concern (under weaker conditions than competing proposals).

Centre for Statistical Methodology Seminar
Time Series Regression Analysis Theme
Friday 23 March 2018, 12:45-2:00pm
LG81, Keppel Street
Case time series: a flexible design for big data epidemiological analyses
Antonio Gasparrini (LSHTM)

Abstract: Biomedical research has been transformed by recent developments in big data technologies. For instance, the collection of health records in linked electronic databases provide information on demographics, health events, medications, and lifestyle factors for large samples of patients. Similarly, portable devices such as mobile phones provide the opportunity to recruit large numbers of participants, and to collect real-time and geo-located individual-level measurements. While these resources offer the possibility to answer research questions that could not be feasibly addressed using traditional studies, they require innovative analytical approaches. This talk will illustrate a novel analytical design called case time series. This study design offers an adaptable framework that combines the individual-level setting and ability to control for confounders of case-only methods, with the flexibility and temporal structure of time series models. It represents a general tool, applicable in different research areas for investigating short-term associations with environmental exposures, clinical conditions, or medications. The case time series design is suitable for the analysis of highly-informative big data resources, particularly those providing individual profiles with longitudinal measures of health outcomes and time-varying predictors.

Centre for Statistical Methodology Seminar
Survival Analysis Theme
Friday 27 April 2018, 12:45-2:00pm
LG9, Keppel Street
Dynamic prediction in fertility
Nan van Geloven (University of Leiden)

Abstract: Dynamic prediction of time-to-pregnancy is challenging due to unobserved patient heterogeneity. As over time couples with better prognoses become pregnant, those remaining behind represent a selected group with relatively poor prognoses. This selection process is not well captured by standard survival models. In this talk, I will present two models that explicitly account for unobserved heterogeneity. In the first, the beta-geometric model is used to make repeated predictions of natural conception over time. In the second, I use frailty models to evaluate the impact of heterogeneity on treatment delay. The predictions from these models can support treatment decisions of subfertility couples during their long and stressful journey towards parenthood.