Target Markdown and {stantargets} for Bayesian model validation pipelines

Will Landau, Senior Research Scientist, presents on using Target Markdown, a system that has all the convenience of Rmarkdown and {stantargets}. Landau covers the workflows, the dilemma of short runtimes or reproducible results, using the pipeline tool, and other challenges. Landau, also describes an example of the Bayesian longitudinal model for clinical trials, ensuring that the model is implemented correctly through calibration. Originally presented at R/Medicine 2021 by Will Landau https://r-medicine.org Main Sections 0:08 Target markdown and {stantargets} for Bayesian model 0:57 Repetition: the overlooked bane of long computation 1:40 Workflows have interconnected steps 1:57 If you change code or data 2:02 The downstream steps are no longer valid 2:27 Dilemma: short runtimes or reproducible results? 2:56 Let a pipeline tool figure out what to return 3:53 Pipeline tools 4:26 Challenge 5:16 Extending targets 5:59 Target factories simplify pipeline construction 7:05 Example: Bayesian model for clinical trials 8:12 Interval-based validation study 8:57 Write the pipeline in Target markdown 9:59 One function to simulate prior predictive data 11:15 Simulations and MCMC with stantargets 12:34 Simple target to convergence diagnostics 13:01 Simple targets for coverage statistics 13:24 Optimal code chunk to run the pipeline 13:49 Optimal code chunks to read the results 14:39 Coverage is nominal More Resources Main Site: https://www.r-consortium.org/ News: https://www.r-consortium.org/news Blog: https://www.r-consortium.org/news/blog Join: https://www.r-consortium.org/about/join Twitter: https://twitter.com/Rconsortium LinkedIn: https://www.linkedin.com/company/r-consortium/