Title: An umbrella test for violations of some instrumental variable assumptions.

Abstract: Point causal effect estimation using instrumental variables requires not only the core instrumental variable assumptions of relevance, independence and exclusions, but also additional parametric assumptions about the instrument-exposure and/or exposure-outcome relationship. In this seminar, I will present a simple test that can be used to assess the plausibility of some of these assumptions jointly in the case of a continuous outcome. The test exploits the fact that, under the assumptions that the instrument is valid and the second-stage model (i.e., the model describing the causal exposure-outcome relationship) is correctly specified, the residuals from the second-stage model would be expected to be homoscedastic with respect to the outcome. Therefore, evidence against the null hypothesis of homoscedasticity would be indicative that at least one of these assumptions is violated. Importantly, the proposed test does not detect all violations of these assumptions, so it is a valid falsification (but not confirmatory) test. 

Biography: Dr Hartwig received his MSc and Doctorate degrees in Epidemiology at Federal University of Pelotas (Brazil), where he currently is an assistant professor in Epidemiology and a permanent researcher at the Postgraduate Program in Epidemiology. Dr Hartwig also holds an honorary research fellow position at Bristol Medical School (University of Bristol, UK). His research interests are related to the field of causal inference in epidemiology, with emphasis on empirical and methodological work on instrumental variable analysis. He is also interested in translation of statistical methods between different fields of epidemiology.

All welcome

In OS6 Oakfield House or Teams 

Join the meeting now  

Meeting ID: 321 686 027 37  

Passcode: nj2UQ7U5