Better Survey Weights Via Regularized Raking with Andy Timm - nyhackr December Meetup

Visit www.nyhackr.org to learn more and follow / nyhackr
Talk Title- Better Survey Weights Via Regularized Raking
Talk Description- Raking is among the most common algorithms for producing survey weights, but it is often opaque what qualities of the resulting weights set are prioritized by the method. This is especially true when practitioners turn to heuristic methods like trimming to improve weights. After reviewing the basic raking algorithm and showing some examples in R, I'll show that survey weighting can also be understood as an optimization problem, one which allows for explicit regularization. In addition to providing a conceptually crisp view of what (vanilla) raking optimizes for, I'll show that this regularized raking (implemented via the rsw python package) can allow for more fine-grained control over weights distributions, and ultimately more accurate weighted estimates. Examples will be drawn from US elections surveys.
Bio- Andy is a data scientist who enjoys using modern statistical and machine learning methods to more efficiently solve social science flavored problems (like survey weighting!). Most recently, he was a Data Science Manager at 605, where he worked on problems spanning causal inference, forecasting, and heterogeneous treatment effect modeling. Before 605, Andy worked in politics in a variety of data and field roles. He holds a masters from NYU Gallatin, and a BA from Macalester College.

[@ricardopietrobon1222]

Outstanding presentation!! Just too bad that the slides were not matching the talk