Learning Bayesian Statistics #1 Bayes, open-source and bioinformatics, with Osvaldo Martin
Oct 8, 2019
Meet Osvaldo Martin, a biologist, physicist, and data scientist discussing Bayesian methods and open-source development. Explore the intersection of bioinformatics and statistics, the influence of Bayesian methods, and the importance of model comparison in statistics and machine learning. Dive into ARVIS for visualizing Bayesian models and hypothetical dinners with scientific minds.
AI Snips
Chapters
Transcript
Episode notes
From Frustration To Fluency Learning Statistics
- Osvaldo describes learning statistics as a repeated struggle of confusion and abandonment until he found a coherent approach.
- He contrasts this with learning to code (Python), which was gradual and satisfying, citing Matplotlib as the turning point that made plotting by code a game changer.
Automate Repetitive Science Tasks With Code
- Learn tools that let you automate repetitive scientific tasks, like plotting with Matplotlib, to save hours and iterate quickly.
- Osvaldo switched to Python after realizing he could script figure generation and reproduce dozens of plot variations instantly.
Bayesian Methods Are Modeling First
- Osvaldo frames Bayesian statistics primarily as modeling, not just inference: you build explicit models that map closely to the problem.
- He values PyMC because its syntax mirrors textbook models, letting one translate equations to code and then run powerful samplers like NUTS.