Measurement Induced Confounding with George Perrett - nyhackr January Meetup

Visit https://www.nyhackr.org to learn more and follow https://twitter.com/nyhackr Talk Title- Measurement Induced Confounding: or the Assumptions Strike Back Talk Description- Motivation, political ideology, and personality are standard confounding variables frequently adjusted for in observational studies across the social and medical sciences. While the role of confounding variables in observational studies has been thoroughly explored, the measurement of latent confounding variables (variables that can not be directly observed but require measurement through self-reports) has largely been overlooked. In this talk, I will present a brief history of causal inference in observational studies, explain why the role of measurement has been ignored, and discuss why this is problematic. No prior knowledge of causal inference or latent variables is assumed. Over the course of the talk, I will introduce all the necessary concepts, and this talk will be appropriate for both technical and non-technical audiences alike.  Bio- I work on developing novel Bayesian Machine Learning Models to estimate causal effects while minimizing parametric assumptions. Currently, I am interested in data analysis where an analyst is faced with multiple-armed treatments, multiple outcomes, or is considering multiple moderation effects. I care deeply about ensuring that novel statistical methods are accessible to practitioners and do not require niche expertise to implement. Prior to my position as a Visiting Assistant Professor I worked as a research scientist here at NYU.