MENU

Fun & Interesting

The Bayesian Workflow: Building a COVID-19 Model, Part 2 (Thomas Wiecki)

PyMC Developers 4,521 5 years ago
Video Not Working? Fix It Now

Speaker: Thomas Wiecki Title: The Bayesian Workflow: Building a COVID-19 Model (Part 2) Video: https://www.youtube.com/watch?v=_DCkJkMji0U Event description: In this tutorial we will build a COVID-19 model from scratch. Discourse Discussion https://discourse.pymc.io/t/the-bayesian-workflow-building-a-covid-19-model-by-thomas-wiecki/6017 ## Timestamps 00:00 Intro - Motivation for Generative Models in Bayesian Statistics 00:50 Agenda 01:08 Addressing Limitations of previous model using Rt 01:31 Current Rt model in use from rt.live 02:38 Delay effects on Rt we should take into account 03:06 Effects of Tracking lockdowns, mask use and social distancing on Rt 03:45 The generative model 04:21 Dataset - US new daily cases 04:42 Build a generative model using a recursive function for new cases 06:03 Translate the recursive loop into a pymc object with theano.scan 07:06 Instantiate model and set parameters 10:33 Follow the bayesian workflow 11:25 Time varying reproduction 12:28 Select prior with time-varying factor (random walk process) 14:41 Update model with new prior 16:00 Prior and Posterior predictive results 17:23 Taking infection delay into account - generation time distribution 18:00 Modeling of generation time distribution of covid 19 19:03 Convolution definition 20:38 Updating generative process model with a convolution 21:50 Prior and Posterior predictive results 23:18 Compare models 23:49 The number of infected - onset delay distribution 24:50 Update model with onset delay effect 26:26 Prior and Posterior predictive results 27:28 Adjusting for number of tests performed 29:35 Results discussion 30:05 Conclusion / Final remarks Speaker bio: - Thomas is the founder of PyMC Labs, a Bayesian consulting firm. - PyMC author - PhD on computational cognitive neuroscience from Brown University - Former VP of data science and head of research at Quantopian Inc: building a team of data scientists to build a hedge fund from a pool of 300k crowd researchers. - Recognized public speaker: keynotes at the Open Data Science Conference (ODSC) & TACC as well as talks at Strata Hadoop & AI, Newsweek AI conference, and various PyData conferences around the world - Runs a data science podcast, blogs about Bayesian statistics, and is an avid Twitter personality. - Data scientist to follow of the year 2015, ODSC Open Source award 2018. Speaker info: -Website: https://twiecki.io -GitHub: https://github.com/twiecki -Twitter: https://twitter.com/twiecki -LinkedIn: https://www.linkedin.com/in/thomas-wiecki-46339244/ Part of PyMCon2020. More details at http://www.pymcon.com #bayesian #statistics #covid19

Comment