This week, members of the Bay Area useR Group (BARUG) celebrated the group’s one hundred and first meetup with beer, pizza and three outstanding presentations at the cancer diagnostics company GRAIL. Pete Mohanty began the evening with the talk Did “Communities in Crisis” Elect Trump?: An Analysis with Kernel Regularized Least Squares. Not only is the Political Science compelling, but the underlying Data Science is top shelf. The bigKRLS package that Pete and his collaborator Robert Shaffer wrote to support their research uses parallel processing and external memory techniques to make the computationally intensive Regularized Least Squares algorithm suitable for fairly large data sets. The following graph from Pete’s presentation gives some idea of the algorithm’s running time.
In the second talk, long time BARUG contributor Earl Hubbell described the production workflow that supports GRAIL’s scientific work. “Rmarkdown, tidy data principles, and the RStudio ecosystem serve as one foundation for [GRAIL’s] reproducible research.” In the evening’s third talk, Ashley Semanskee the Kaiser Family Foundation is using R to automate the annual production of its Employer Health Benefits Survey.
Harlan Harris gave what looks like must have been an outstanding talk on Forecasting Growth with Semiparametric Bayesain Models to the New York Open Statistical Programming Meetup. Harlan’s code is here.
Note that members of the Portland group also helped to organize the first ever CasdadiaRConf conference.