Juha Karvanen (University of Jyväskylä, Department of Mathematics and Statistics), will lead this webinar on the topic « Causal inference with multiple incomplete data sources ».

Assume that you have conducted an experiment where the effect of A on B is studied. An earlier study indicates that B affects C. Can we combine these data sources to estimate the causal effect of A on C?  What if we have only observational data on A and B?  

In this talk, I will present a tool for causal effect identification from multiple data sources. The data sources may suffer from selection bias and missing data. The do-search algorithm (implemented in the R package dosearch) can solve a wide range of causal and non-causal identifiability problems. The examples presented include combining observations and experiments, identification in missing not at random scenarios and causal inference under the case-control design. In an illustration, data from NHANES 2013-2016 surveys and results from a published meta-analysis are combined to estimate the reduction in average systolic blood pressure under an intervention where the use of table salt is discontinued.