Learning Bayesian Statistics

🎙️ Episode 121 is live! 🎧 https://lnkd.in/gNUNFX3c In this LBS episode, Alexandre Andorra and Nathaniel Forde explore: 🧠 How CFA validates theoretical constructs with a solid theoretical structure 📊The flexibility of Bayesian approaches in modelling complex relationships ✅ Key steps in model validation: global/local fit, sensitivity analysis, and resolving divergences ⚖️ Balancing model complexity, theoretical relevance, and real-world data fitting 🔍 Using factor analysis and survey data to uncover causal relationships and complex phenomena 🗣️ Why philosophy sharpens reasoning and communication is vital for data scientists #LearningBayesianStatistics #BayesianModeling #Psychometrics #CausalInference

📄Check out this paper submitted to the 2025 MIT Sloan Sports Analytics Conference: "Unveiling True Talent: The Soccer Factor Model for Skill Evaluation." The SFM analyzes over 33,000 matches (2000/21–2023/24) to separate individual skill from team effects, delivering better forecasts 📈 and fair comparisons ⚖️ across players. It even tackles debates like the GOAT of the 21st century! 🐐 Maximilian Göbel Alexandre Andorra #LearningBayesianStatistics #SportsAnalytics #GaussianProcesses

Our PyData NYC tutorial with Christopher Fonnesbeck on Mastering #GaussianProcesses with PyMC is now online 🥳 If you missed it live and want to learn more about GPs in practice, come check it out! We cover: 1. Introduction to Probabilistic Modeling (10 min)   - Overview of modeling complex systems using Gaussian distributions. 2. What is a Gaussian Process? (15 min)   - Explanation of Gaussian Processes, their features, and properties.   - Discussion of the flexibility of GPs in capturing non-linear relationships. 3. Building a Gaussian Process Model in PyMC (10 min)   - How to choose and customize covariance functions to suit different problems. 4. Marginal, Latent, and Sparse GPs (10 min)   - Different classes of GPs and their properties. 5. Fast Gaussian Processes: HSGP Approximation (10 min)   - Introduction to the HSGP approximation and how it scales GPs to large datasets. 6. Sports Analytics Case Study (40 min)   - Application of GPs to a real-world sports analytics problem.   - Step-by-step walkthrough of modeling and prediction using GPs in PyMC. Feel free to comment below with questions, editions and suggestions! Enjoy & PyMCheers 🖖 Video: https://lnkd.in/dPee5PYp GitHub repo: https://lnkd.in/ddK28JDA

📢 Episode 120 is Live! 🎧 https://lnkd.in/gbqsPuNC 🎙️ In this LBS episode, Alexandre Andorra, Elizaveta Semenova, and Chris Wymant explore the intersection of epidemiology, Bayesian statistics, and data science! Key Highlights: 🔬 Epidemiology connects health insights across scales, while Bayesian statistics quantifies uncertainty and links models to data. 📈 Advances in data collection and international collaboration, especially post-COVID, are transforming the field. 🤝 Effective research thrives on collaboration between domain experts, statisticians, and machine learning innovations. 💻 Tackling challenges in coding, communication, and understanding data limitations is essential for progress. #LearningBayesianStatistics #Epidemiology #DataScience

🎙️We’re bringing your favourite Bayesian podcast to Bluesky Social! 📊Join us for insights, discussions, and more on Bayesian reasoning and statistics. Follow us if you're into stats, data, or just curious to learn something new. See you there!👋 https://lnkd.in/gxm3NK4u #LearningBayesianStatistics #Bayesian #Podcast

𝐂𝐨𝐢𝐧𝐬 𝐇𝐚𝐯𝐞 𝐚 𝐒𝐞𝐜𝐫𝐞𝐭 𝐁𝐢𝐚𝐬 𝐚𝐧𝐝 𝐁𝐚𝐲𝐞𝐬𝐢𝐚𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 𝐉𝐮𝐬𝐭 𝐄𝐱𝐩𝐨𝐬𝐞𝐝 𝐈𝐭! 🎲 Do you think a coin flip is the ultimate fair decider? Not so fast. After 350,757 𝐟𝐥𝐢𝐩𝐬, Bayesian analysis revealed some surprising truths about this iconic act of “randomness” 🎲 𝐒𝐚𝐦𝐞-𝐒𝐢𝐝𝐞 𝐁𝐢𝐚𝐬: Coins land on the 𝐬𝐚𝐦𝐞 𝐬𝐢𝐝𝐞 they started 50.8% 𝐨𝐟 𝐭𝐡𝐞 𝐭𝐢𝐦𝐞. The 𝐃𝐇𝐌 𝐦𝐨𝐝𝐞𝐥 of rotational dynamics explains this subtle tilt. ⚖️ 𝐇𝐞𝐚𝐝𝐬 𝐯𝐬. 𝐓𝐚𝐢𝐥𝐬? 𝐒𝐭𝐢𝐥𝐥 𝐅𝐚𝐢𝐫: The 50/50 split holds steady for heads vs. tails. Coins may wobble, but they don’t play favourites here. 📈 𝐏𝐫𝐚𝐜𝐭𝐢𝐜𝐞 𝐌𝐚𝐤𝐞𝐬 𝐈𝐭 𝐅𝐚𝐢𝐫-𝐞𝐫: Novice tossers introduce more wobble, increasing the same-side bias. With 10,000+ flips, the bias diminishes as flippers approach the mythical wobble-less toss. 😎 What’s cooler? 𝐁𝐚𝐲𝐞𝐬𝐢𝐚𝐧 𝐡𝐢𝐞𝐫𝐚𝐫𝐜𝐡𝐢𝐜𝐚𝐥 𝐦𝐨𝐝𝐞𝐥𝐬 helped disentangle the bias, revealing fascinating individual differences among flippers. The takeaway for high-stakes flips? 𝐂𝐨𝐧𝐜𝐞𝐚𝐥 𝐭𝐡𝐞 𝐬𝐭𝐚𝐫𝐭𝐢𝐧𝐠 𝐬𝐢𝐝𝐞 if you want a truly fair toss. Explore the full study and data 👇 🪙 https://lnkd.in/dfMymF3d #LearningBayesianStatistics #BayesianAnalysis #DataScience

🔥 *ALL* models make prior assumptions - Bayesian just gives you control of them. ➡️  By choosing a non-Bayesian model you’re implicitly choosing a prior. ➡️  Most models implicitly assume the equivalent to an ‘improper flat prior’ ➡️  This means that all possibilities are equally likely before seeing the data. 💡 You can choose this prior in Bayesian models too. But given the choice you almost never choose it! This simple example of prior predictive plots of predicting weight from height demonstrates this. The Bayesian prior is only 'subjective' in that we're reducing the chance of ridiculous or impossible values, and ensuring the slope will be non-negative. This is a common sense starting position, gives us regularisation, and is why Bayesian approaches do so much better with less data. They have either common sense or expert knowledge built in in advance of even seeing the data.

🐘 We’re on Mastodon! Follow us for the good stuff 😉 👉 https://lnkd.in/gNzAWXnd #LearningBayesianStatistics #Mastodon #Community

📢 Episode 119 is live! 🎧 https://lnkd.in/ggd2jgKi 🎙️ In this 𝑳𝑩𝑺 episode, Alexandre Andorra and Robert Kubinec discuss corruption research, the challenges of measuring it, and survey techniques like randomized response. They delve into the power of Bayesian methods for latent variables, insights from Kubinec's novel, the importance of the Stan community, and the potential of real-time online surveys for advancing social science research. 🔑 Key Takeaways: 📉 Corruption is difficult to measure due to its hidden nature and challenges in gathering truthful data.  📊 Bayesian methods are useful for estimating latent variables and using prior information.  📋 Techniques like randomized response and real-time online surveys improve data collection for sensitive topics.  ✍️ Writing fiction can support creativity and clarity in academic work.  #LearningBayesianStatistics #BayesianAnalysis

Seven years ago, I dove into the world of #BayesianStatistics, soaking up every tutorial I could find. A pivotal moment was Christopher Fonnesbeck's #GaussianProcesses workshop at PyData NYC, which ignited my passion and set the foundation for my growth. Last week, I had the incredible honor of *teaching* an Advanced GPs tutorial in PyMC... alongside the very same Chris Fonnesbeck 🤩 It was surreal to come full circle and contribute to a field that has given me so much. This path has been paved with sweat, tears, and relentless perseverance, but the rewards are immeasurable—especially when guided by inspiring mentors 🙏 A massive thank you to Chris for his unwavering dedication to the community and to the entire PyMC team for their continuous support and innovation. Your efforts empower learners like me to push boundaries and achieve our goals. Here’s to the power of mentorship, community, and never stopping the quest for knowledge! 🚀✨ As if it weren't enough, I got the chance to meet fantastic people at the Flatiron Institute, great folks like Nathaniel Haines, and record live Learning Bayesian Statistics episodes with the great Aaron MacNeil and Will Dean -- coming very soon in your feed 😉 A huge thank you to PyData Global, NumFOCUS and Tomara Youngblood for their trust and organization 🙏 See you next one! Live Bayes & Prosper 🖖