Mendelian Randomisation for Causal Inference in Observational Epidemiology Nuala Sheehan, University of Leicester. Assessing the effect of some modifiable exposure on a disease from observational data is often important for informing health intervention policies. To be confident that a particular intervention will actually be useful, it is necessary to ascertain that an observed association or correlation between the exposure and disease means that the risk factor is causal for the disease. Inferring causality from observational data is difficult as it is not always clear which of two associated variables is the cause, which the effect, or whether both are common effects of a third unobserved variable, or confounder. A possible approach to testing for or estimating causal effects when confounding is believed to be present but not fully understood is based on the method of instrumental variables which are widely used in econometrics. It is known under the name of Mendelian randomisation, if the instrument is a genetic variant, and has received a lot of attention in the epidemiological literature recently. Testing for the presence of a causal effect is generally straightforward but point estimates are only obtainable under additional parametric and distributional assumptions. Such assumptions cannot be tested directly from the observable quantities and thus must be justified from background or application-specific knowledge. Different IV estimators require different assumptions and target different causal effects, some of which are not necessarily meaningful in an observational setting. Problems particularly arise when the outcome of interest is a binary indicator of disease status, for example, although there are special cases where these can be satisfactorily addressed. A Bayesian approach could also be considered but also requires assumptions that have to be justified.