Learning Bayesian Statistics

In times of crisis, designing an efficient policy response is paramount. In case of natural disasters or pandemics, it can even determine the difference between life and death for a substantial number of people. But precisely, how do you design such policy responses, making sure that risks are optimally shared, people feel safe enough to reveal necessary information, and stakeholders commit to the policies?

That’s where a field of economics, industrial organization (IO), can help, as Shosh Vasserman will tell us in this episode. Shosh is an assistant professor of economics at the Stanford Graduate School of Business. Specialized in industrial organization, her interests span a number of policy settings, such as public procurement, pharmaceutical pricing and auto-insurance.

Her work leverages theory, empirics and modern computation (including the Stan software!) to better understand the equilibrium implications of policies and proposals involving information revelation, risk sharing and commitment. 

In short, Shoshana uses theory and data to study how risk, commitment and information flows interplay with policy design. And she does a lot of this with… Bayesian models! Who said Bayes had no place in economics?

Prior to Stanford, Shoshana did her Bachelor’s in mathematics and economics at MIT, and then her PhD in economics at Harvard University.

This was a fascinating conversation where I learned a lot about Bayesian inference on large scale random utility logit models, socioeconomic network heterogeneity and pandemic policy response — and I’m sure you will too!

In a few days, a consequential election will take place, as citizens of the United States will go to the polls and elect their president — in fact they already started voting. You probably know a few forecasting models that try to predict what will happen on Election Day — who will get elected, by how much and with which coalition of States?

But how do these statistical models work? How do you account for the different sources of uncertainty, be it polling errors, unexpected turnout or media events? How do you model covariation between States? How do you even communicate the model’s results and afterwards assess its performance? To talk about all this, I had the pleasure to talk to Andrew Gelman and Merlin Heidemanns.

Andrew was already on episode 20, to talk about his recent book with Jennifer Hill and Aki Vehtari, “Regression and Other Stories”. He’s a professor of statistics and political science at Columbia University and works on a lot of topics, including: why campaign polls are so variable while elections are so predictable, the statistical challenges of estimating small effects, and methods for surveys and experimental design.

Merlin is a PhD student in Political Science at Columbia University, and he specializes in political methodology. Prior to his PhD, he did a Bachelor's in Political Science at the Freie Universität Berlin.

I hope you’ll enjoy this episode where we dove into the Bayesian model they helped develop for The Economist, and talked more generally about how to forecast elections with statistical methods, and even about the incentives the forecasting industry has as a whole.

I don’t know about you, but I’m starting to really miss traveling and just talking to people without having to think about masks, social distance and activating the covid tracking app on my phone. In the coming days, there is one event that, granted, won’t make all of that disappear, but will remind me how enriching it is to meet new people — this event is PyMCon, the first-ever conference about the PyMC ecosystem! To talk about the conference format, goals and program, I had the pleasure to host Ravin Kumar and Quan Nguyen on the show.

Quan is a PhD student in computer science at Washington University in St Louis, USA, researching Bayesian machine learning and one of the PyMCon program committee chairs. He is also the author of several programming books on Python and scientific computing.

Ravin is a core contributor to Arviz and PyMC, and is leading the PyMCon conference. He holds a Bachelors in Mechanical Engineering and a Masters in Manufacturing Engineering. As a Principal Data Scientist he has used Bayesian Statistics to characterize and aid decision making at organizations like SpaceX and Sweetgreen. Ravin is also currently co-authoring a book with Ari Hartikainen, Osvaldo Martin, and Junpeng Lao on Bayesian Statistics due for release in February.

We talked about why they became involved in the conference, parsed through the numerous, amazing talks that are planned, and detailed who the keynote speakers will be… So, If you’re interested the link to register is in the show notes, and there are even two ways to get a free ticket: either by applying to a diversity scholarship, or by being a community partner, which is anyone or any organization working towards diversity and inclusion in tech — all links are in the show notes.

Have you watched the series « The English Game », on Netflix? Well, I think you should — it’s a fascinating dive into how football went from an aristocratic to a popular sport in the late 19th century England. Today it is so popular that it became a valuable business to do statistics on the game and its players!

To talk about that, I invited Kevin Minkus on the show — he’s a data scientist and soccer fan living in Philadelphia. Kevin’s currently working at Monetate on ecommerce problems, and prior to Monetate he worked on property and casualty insurance pricing.

He spends a lot of his spare time working on problems in football analytics and is a contributor at American Soccer Analysis, a website and podcast dedicated to… football made or played in the US (or “soccer”, as they say over there). Kevin is responsible for some of their data management and devops, and he recently wrote a guide to football analytics for the Major League Soccer’s website, entitled « Soccer Analytics 101 ».

To be honest, I had a great time talking for one hour about two of my passions — football and stats! Soooo, maybe 2020 isn’t that bad after all… Ow, and beyond football, Kevin is also into the digital humanities, web development, 3D animation, machine learning, and… the bassoon!

Do you know what proteins are, what they do and why they are useful? Well, be prepared to be amazed! In this episode, Seth Axen will tell us about the fascinating world of protein structures and computational biology, and how his work of Bayesian modeler fits into that!

Passionate about mathematics and statistics, Seth is finishing a PhD in bioinformatics at the Sali Lab of the University of California, San Francisco (UCSF). His research interests span the broad field of computational biology: using computer science, mathematics, and statistics to understand biological systems. His current research focuses on inferring protein structural ensembles. 

Open source development is also very dear to his heart, and indeed he contributes to many open source packages, especially in the Julia ecosystem. In particular, he develops and maintains ArviZ.jl, the Julia port of ArviZ, a platform-agnostic python package to visualize and diagnose your Bayesian models. Seth will tell us how he became involved in ArviZ.jl, what its strengths and weaknesses are, and how it fits into the Julia probabilistic programming landscape.

Ow, and as a bonus, you’ll discover why Seth is such a fan of automatic differentiation, aka « autodiff » — I actually wanted to edit this part out but Seth strongly insisted I kept it. Just kidding of course — or, am I… ?

If you’ve studied at a business school, you probably didn’t attend any Bayesian stats course there. Well this isn’t like that in every business schools! Elea McDonnel Feit does integrate Bayesian methods into her teaching at the business school of Drexel University, in Philadelphia, US. 

Elea is an Assistant Professor of Marketing at Drexel, and in this episode she’ll tell us which methods are the most useful in marketing analytics, and why.

Indeed, Elea develops data analysis methods to inform marketing decisions, such as designing new products and planning advertising campaigns. Often faced with missing, unmatched or aggregated data, she uses MCMC sampling, hierarchical models and decision theory to decipher all this.

After an MS in Industrial Engineering at Lehigh University and a PhD in Marketing at the University of Michigan, Elea worked on product design at General Motors and was most recently the Executive Director of the Wharton Customer Analytics Initiative.

Thanks to all these experiences, Elea loves teaching marketing analytics and Bayesian and causal inference at all levels. She even wrote the book R for Marketing Research and Analytics with Chris Chapman, at Springer Press.

In summary, I think you’ll be pretty surprised by how Bayesian the world of marketing is…

If, like me, you’ve been stuck in a 40 square-meter apartment for two months, you’re going to be pretty jealous of Avi Bryant. Indeed, Avi lives on Galiano Island, Canada, not very far from Vancouver, surrounded by forest, overlooking the Salish Sea. 

In this natural and beautiful — although slightly deer-infested — spot, Avi runs The Gradient Retreat Center, a place where writers, makers, and code writers can take a week away from their regular lives and focus on creative work. But it’s not only to envy him that I invited Avi on the show — it’s to talk about Bayesian inference in Scala, prior elicitation, how to deploy Bayesian methods at scale, and how to enable Bayesian inference for engineers. 

While working at Stripe, Avi wrote Rainier, a Bayesian inference framework for Scala. Inference is based on variants of the Hamiltonian Monte Carlo sampler, and the implementation is similar to, and targets the same types of models as both Stan and PyMC3. As Avi says, depending on your background, you might think of Rainier as aspiring to be either "Stan, but on the JVM", or "TensorFlow, but for small data".

In this episode, Avi will tell us how Rainier came into life, how it fits into the probabilistic programming landscape, and what its main strengths and weaknesses are.

I bet you heard a lot about epidemiological compartmental models such as SIR in the last few months? But what are they exactly? And why are they so useful for epidemiological modeling? 

Elizaveta Semenova will tell you why in this episode, by walking us through the case study she recently wrote with the Stan team. She’ll also tell us how she used Gaussian Processes on spatio-temporal data, to study the spread of Malaria, or to fit dose-response curves in pharmaceutical tests. 

And finally, she’ll tell us how she used Bayesian neural networks for drug toxicity prediction in her latest paper, and how Bayesian neural nets behave compared to classical neural nets. Ow, and you’ll also learn an interesting link between BNNs and Gaussian Processes…

I know: Liza works on _a lot_ of projects! But who is she? Well, she’s a postdoctorate in Bayesian Machine Learning at the pharmaceutical company AstraZeneca, in Cambridge, UK. 

Elizaveta did her masters in theoretical mathematics in Moscow, Russia, and then worked in financial services as an actuary in various European countries. She then did a PhD in epidemiology at the University of Basel, Switzerland. This is where she got interested in health applications – be it epidemiology, global health or more small-scale biological questions. But she’ll tell you all that in the episode ;)

Once upon a time, there was an enchanted book filled with hundreds of little plots, applied examples and linear regressions — the prettiest creature that was ever seen. Its authors were excessively fond of it, and its readers loved it even more. This magical book had a nice blue cover made for it, and everybody aptly called it « Regression and other Stories »!

As every good fairy tale, this one had its share of villains — the traps where statistical methods fall and fail you; the terrible confounders, lurking in the dark; the ill-measured data that haunt your inferences! But once you defeat these monsters, you’ll be able to think about, build and interpret regression models.

This episode will be filled with stories — stories about linear regressions! Here to narrate these marvelous statistical adventures are Andrew Gelman, Jennifer Hill and Aki Vehtari — the authors of the brand new Regression and other Stories.

Andrew is a professor of statistics and political science at Columbia University. Jennifer is a professor of applied statistics at NYU. She develops methods to answer causal questions related to policy research and scientific development. Aki is an associate professor in computational probabilistic modeling at Aalto University, Finland.

In this episode, they tell us why they wrote this book, who it is for and they also give us their 10 tips to improve your regression modeling! We also talked about the limits of regression and about going to Mars…

Do you know Turing? Of course you do! With Soss and Gen, it’s one of the blockbusters to do probabilistic programming in Julia. And in this episode Cameron Pfiffer will tell us all about it — how it came to life, how it fits into the probabilistic programming landscape, and what its main strengths and weaknesses are.

Cameron did some Rust, some Python, but he especially loves coding in Julia. That’s also why he’s one of the core-developers of Turing.jl. He’s also a PhD student in finance at the University of Oregon and did his master’s in finance at the University of Reading. His interests are pretty broad, from cryptocurrencies, algorithmic and high-frequency trading, to AI in financial markets and anomaly detection – in a nutshell he’s a fan of topics where technology is involved.

As he’s the first economist to come to the show, I also asked him how Bayesian the field of economics is, why he thinks economics is quite unique among the social sciences, and how economists think about causality — I later learned that this topic is pretty controversial!

I hope you’re all safe! Some of you also asked me if I had set up a Patreon so that they could help support the show, and that’s why I’m sending this short special episode your way today. I had thought about that, but I wasn’t sure there was a demand for this. Apparently, there is one — at least a small one — so, first, I wanna thank you and say how grateful I am to be in a community that values this kind of work!

How do you design a good experimental study? How do you even know that you’re asking a good research question? Moreover, how can you align funding and publishing incentives with the principles of an open source science?

Let’s do another “big picture” episode to try and answer these questions! You know, these episodes that I want to do from time to time, with people who are not from the Bayesian world, to see what good practices there are out there. The first one, episode 15, was focused on programming and python, thanks to Michael Kennedy. 

In this one, you’ll meet Daniel Lakens. Daniel is an experimental psychologist at the Human-Technology Interaction group at Eindhoven University of Technology, in the Netherlands. He’s worked there since 2010, when he received his PhD in social psychology. 

His research focuses on how to design and interpret studies, applied meta-statistics, and reward structures in science. Daniel loves teaching about research methods and about how to ask good research questions. He even crafted free Coursera courses about these topics. 

A fervent advocate of open science, he prioritizes scholar articles review requests based on how much the articles adhere to Open Science principles. On his blog, he describes himself as ‘the 20% Statistician’. Why? Well, he’ll tell you in the episode…

Have you already encountered a model that you know is scientifically sound, but that MCMC just wouldn’t run? The model would take forever to run — if it ever ran — and you would be greeted with a lot of divergences in the end. Yeah, I know, my stress levels start raising too whenever I hear the word « divergences »…

Well, you’ll be glad to hear there are tricks to make these models run, and one of these tricks is called re-parametrization — I bet you already heard about the poorly-named non-centered parametrization?

Well fear no more! In this episode, Maria Gorinova will tell you all about these model re-parametrizations! Maria is a PhD student in Data Science & AI at the University of Edinburgh. Her broad interests range from programming languages and verification, to machine learning and human-computer interaction. 

More specifically, Maria is interested in probabilistic programming languages, and in exploring ways of applying program-analysis techniques to existing PPLs in order to improve usability of the language or efficiency of inference.

As you’ll hear in the episode, she thinks a lot about the language aspect of probabilistic programming, and works on the automation of various “tricks” in probabilistic programming: automatic re-parametrization, automatic marginalization, automatic and efficient model-specific inference.

As Maria also has experience with several PPLs like Stan, Edward2 and TensorFlow Probability, she’ll tell us what she thinks a good PPL design requires, and what the future of PPLs looks like to her.

A librarian, a philosopher and a statistician walk into a bar — and they can’t find anybody to talk to; nobody seems to understand what they are talking about. Nobody? No! There is someone, and this someone is Will Kurt! 

This is it folks! This is the first of the special episodes I want to do from time to time, to expand our perspective and get inspired by what’s going on elsewhere. The guests will not come directly from the Bayesian world, but will still be related to science or programming.

For the first episode of the kind, I had the chance to chat with Michael Kennedy! Michael is not only a very knowledgeable and respected member of the Python community, he’s also the founder and host of Talk Python To Me, the most popular Python podcast. He’s the founder and chief author at Talk Python Training, where he develops many Python developer online courses. 

And before that, Michael was a professional software trainer for over 10 years – he has taught numerous developers throughout the world! But Michael is not only an entrepreneur and teacher – he’s also a father, a husband, and a proud inhabitant of Portland, OR! 

As you’ll hear, our conversation spanned a large array of topics — the role of Python in science and research; how it came to be so important in data science, and why; what are Python’s threats and weaknesses and how it should evolve to not become obsolete. Michael also has interesting thoughts on the role of programming in education and how it relates to geometry — but I’ll let you discover that one by yourself…

I bet you love penguins, right? The same goes for koalas, or puppies! But what about sharks? Well, my next guest loves sharks — she loves them so much that she works a lot with marine biologists, even though she’s a statistician! 

Vianey Leos Barajas is indeed a statistician primarily working in the areas of statistical ecology, time series modeling, Bayesian inference and spatial modeling of environmental data. Vianey did her PhD in statistics at Iowa State University and is now a postdoctoral researcher at North Carolina State University.

In this episode, she’ll tell us what she’s working on that involves sharks, sheep and other animals! Trying to model animal movements, Vianey often encounters the dreaded multimodal posteriors. She’ll explain why these can be very tricky to estimate, and why ecological data are particularly suited for hidden Markov models and spatio-temporal models — don’t worry, Vianey will explain what these models are in the episode!

How is Julia doing? I’m talking about the programming language, of course! What does the probabilistic programming landscape in Julia look like? What are Julia’s distinctive features, and when would it be interesting to use it?

To talk about that, I invited Chad Scherrer. Chad is a Senior Research Scientist at RelationalAI, a company that uses Artificial Intelligence technologies to solve business problems.

Coming from a mathematics background, Chad did his PhD at Indiana University of Bloomington and has been working in statistics and data science for a decade now. Through this experience, he’s been using and developing probabilistic programming languages – so he’s familiar with python, R, PyMC, Stan and all the blockbusters of the field. 

But since 2018, he’s particularly interested in Julia and developed Soss, an open-source lightweight probabilistic programming package for Julia. In this episode, he’ll tell us why he decided to create this package, and which choices he made that made Soss what it is today. But we’ll also talk about other projects in Julia, like Turing or Gen for instance.

Do you know Google Summer of Code? It’s a time of year when students can contribute to open-source software by developing and adding much needed functionalities to the open-source package of their choice. And Demetri Pananos did just that.

He did it in 2019 with PyMC3, for which he developed the API for ordinary differential equations. In this episode, he’ll tell us why and how he did that, what he learned from the experience, and what the strengths and weaknesses of the API are in his opinion.

Demetri is a Ph.D candidate in Biostatistics at Western University, in Ontario, Canada. His research interests surround machine learning and Bayesian statistics for personalized medicine. He earned his Master’s in Applied Mathematics from The University of Waterloo and is a firm believer in open science, interdisciplinary collaboration, and reproducible research. 

Other than that, he loves plotting data and drinking IPA beer – well, who doesn’t?”

I bet you already heard about hierarchical models, or multilevel models, or varying-effects models — yeah this type of models has a lot of names! Many people even turn to Bayesian tools to build _exactly_ these models. But what are they? How do you build and use a hierarchical model? What are the tricks and classical traps? And even more important: how do you _interpret_ a hierarchical model?

In this episode, Thomas Wiecki will come to the rescue and explain what multilevel models are, how to build them, what their powers are… but also why you should be very careful when building them…

Does the name Thomas Wiecki ring a bell? Probably because he’s the host and creator of the PyData Deep Dive Podcast, where he interviews open-source contributors from the Python and Data Science worlds! Thomas is also the VP of Data Science at Quantopian, a crowd-sourced quantitative investment firm that encourages people everywhere to write investment algorithms.

Finally, Thomas is a longtime Bayesian and core-developer of PyMC3, a fantastic python package to do probabilistic programming in Python. On his blog, he publishes tutorial articles and explores new ideas such as Bayesian Deep Learning. Caring a lot about open-source software sustainability, he puts all he’s up to on his Patreon page, that you’ll find in the show notes.

How do you handle your MCMC samples once your Bayesian model fit properly? Which diagnostics do you check to see if there was a computational problem? And isn’t that nice when you have beautiful and reliable plots to complement your analysis and better understand your model?

I know what you think: plotting can be long and complicated in these cases. Well, not with ArviZ, a platform-agnostic package to do exploratory analysis of your Bayesian models. And in this episode, Ari Hartikainen will tell you why.

Ari is a data-scientist in geophysics and a researcher at the Department of Civil Engineering of Aalto University in Finland. He mainly works on geophysics, Bayesian statistics and visualization. 

Ari’s also a prolific open-source contributor, as he’s a core-developer of the popular Stan and ArviZ libraries. He’ll tell us how PyStan interacts with ArviZ, what he thinks ArviZ most useful features are, and which common difficulties he encounters with his models and data.

Have you always wondered what dark matter is? Can we even see it — let alone measure it? And what would discover it imply for our understanding of the Universe?

In this episode, we’ll take look at the cosmos with Maggie Lieu. She’ll tell us what research in astrophysics is made of, what model she worked on at the European Space Agency, and how Bayesian the world of space science is.

Maggie Lieu did her PhD in the Astronomy & Space Department of the University of Birmingham. She’s now a Research Fellow of Machine Learning & Cosmology at the University of Nottingham and is working on projects in preparation for Euclid, a space-based telescope whose goal is to map the dark Universe and help us learn about the nature of dark matter and dark energy.

In a nutshell, she tries to help us better understand the entire cosmos. Even more amazing, she uses the Stan library and applies Bayesian statistical methods to decipher her astronomical data! But Maggie is not just a Bayesian astrophysicist: she also loves photography and rock-climbing!

What is it like using Bayesian tools when you’re a software engineer or computer scientist? How do you apply these tools in the online ad industry? 

More generally, what is Bayesian thinking, philosophically? And is it really useful in every day life? Because, well you can’t fire up MCMC each time you need to make a quick decision under uncertainty… So how do you do that in practice, when you have at most a pen and paper?

In this episode, you’ll hear Max Sklar’s take on these questions. Max is a software engineer with a focus on machine learning and Bayesian inference. Now working at Foursquare’s innovation lab, he recently led the development of a causality model for Foursquare’s Ad Attribution product and taught a course on Bayesian Thinking at the Lviv Data Science Summer School.

Max is also an open-source enthusiast and a fellow podcaster – he’s the host of the Local Maximum podcast, where you can hear every week about the latest trends in AI, machine learning and technology from an engineering perspective.

Ow, and if you liked the movie « Her », with Joaquin Phoenix, well you’re in for a treat at the end of this episode…

You can’t study psychology up until your PhD and end-up doing very mathematical and computational data science at Google right? It’s too hard of a U-turn — some would even say it’s NUTS, just because they like bad puns… Well think again, because Junpeng Lao did just that!

Before doing data science at Google, Junpeng was a cognitive psychology researcher at the University of Fribourg, Switzerland. Working in Python, Matlab and occasionally in R, Junpeng is a prolific open-source contributor, particularly to the popular TensorFlow and PyMC3 libraries. He also maintains the PyMC Discourse on his free time, where he amazingly answers all kinds of various and very specific questions!

In this episode, he’ll tell you what the core characteristics of TensorFlow Probability are, and when you would use TFP instead of another probabilistic programming framework, like Stan or PyMC3. He’ll also explain why PyMC4 will be based on TensorFlow Probability itself, and what future contributions he has in mind for these two amazing libraries. Finally, Junpeng will share with you his workflow for debugging a model, or just for better understanding your models.

If you’re there, it’s probably because you’re interested in Bayesian inference, right? But don’t you feel lost sometimes when building a  model? Or you ask yourself why what you’re trying to do is so damn hard… and you conclude that YOU are the problem, that YOU must be doing something wrong!

Well, rest assured, as you’ll hear from Michael Betancourt himself: it’s hard for everybody! That’s why over the years  he developed and tries to popularize what he calls a « principled  Bayesian workflow » — in a nutshell, think about what could have  generated your data; and always question default settings!

With  that workflow, you’ll probably feel less alone when modeling, but expect to fail often. That’s ok — as Michael says: if you don’t fail, you don’t learn!

Who is Michael Betancourt you ask? He is a physicist and  statistician, whose research focuses on the development of robust statistical workflows, computational tools, and pedagogical resources that help bridge the gap between statistical theory and scientific practice.

Michael works a lot on differential geometry and  probability theory, and he often lives in high-dimensional spaces, where  he meets with a good friend of his -- Hamiltonian Monte Carlo. Then, you won’t be surprised to learn that Michael is one of the core developers of the seminal probabilistic programming language Stan.

I have two questions for you: Are you a self-learner? Then how do  you stay up to date? What should you focus on if you’re a beginner, or if you’re more advanced?
And here is my second question: Are you working in biomedicine? And if you do, are you using Bayesian tools?  Then how do you get your co-workers more used to posterior distributions  than p-values? In other words, how do you change behaviors in a large organization?

In this episode, Eric Ma will answer all these  questions and even tell us his favorite modeling techniques, which  problems he encountered with these models, and how he solved them. He’ll also share with us the software-engineering workflow he uses at Novartis to share his work with colleagues.

Eric is a data  scientist at the Novartis Institutes for Biomedical Research, where he focuses on Bayesian statistical methods to make medicines for patients. Eric is also a prolific open source developer: he led the development of pyjanitor, an API for cleaning data in Python, and nxviz, a  visualization package for NetworkX. He also contributes to PyMC3, matplotlib and bokeh.

This is « Learning Bayesian Statistics », episode 5, recorded October 21, 2019.

What do neurodegenerative diseases, gerrymandering and ecological inference all have in common? Well, they can all be studied with Bayesian methods — and that’s exactly what Karin Knudson is doing.

In this episode, Karin will share with us the vital and essential work she does to understand aspects of neurodegenerative diseases. She’ll also tell us more about computational neuroscience and Dirichlet processes — what they are, what they do, and when you should use them.

Karin did her doctorate in mathematics, with a focus on compressive sensing and computational neuroscience at the University of Texas at Austin. Her doctoral work included applying hierarchical Dirichlet processes in the setting of neural data and focused on one-bit compressive sensing and spike-sorting.

Formerly the chair of the math and computer science department of Phillips Academy Andover, she started a postdoc at Mass General Hospital and Harvard Medical in Fall 2019. Most importantly, rock climbing and hiking have no secrets for her!

How can you use Bayesian tools and optimize your models in industry? What are the best ways to communicate and visualize your models with non-technical and executive people? And what are the most common pitfalls?

In this episode, Colin Carroll will tell us how he did all that in finance and the airline industry. He’ll also share with us what the future of probabilistic programming looks like to him.

You already heard from Colin two weeks ago — so, if you didn’t catch this episode, go back in your feed’s history and enjoy the first part! 

As a reminder, Colin is a machine learning researcher and software engineer who’s notably worked on modeling risk in the airline industry and building NLP-powered search infrastructure for finance. He’s also an active contributor to open source, particularly to the popular PyMC3 and ArviZ libraries.

When speaking about Bayesian statistics, we often hear about « probabilistic programming » — but what is it? Which languages and libraries allow you to program probabilistically? When is Stan, PyMC, Pyro or any other probabilistic programming language most appropriate for your project? And when should you even use Bayesian libraries instead of non-bayesian tools, like Statsmodels or Scikit-learn?

Colin Carroll will answer all these questions for you. Colin is a machine learning researcher and software engineer who’s notably worked on modeling risk in the airline industry and building NLP-powered search infrastructure for finance. He’s also an active contributor to open source, particularly to the popular PyMC3 and ArviZ libraries.

Having studied geometric measure theory at Rice University, Colin was bound to walk in the woods with Pete the pup – who was there when we recorded by the way – and to launch balloons into near-space in his spare time.

When are Bayesian methods most useful? Conversely, when should you NOT use them? How do you teach them? What are the most important skills to pick-up when learning Bayes? And what are the most difficult topics, the ones you should maybe save for later?

In this episode, you’ll hear Chris Fonnesbeck answer these questions from the perspective of marine biology and sports analytics. Chris is indeed the New York Yankees’ senior quantitative analyst and an associate professor at Vanderbilt University School of Medicine. 

He specializes in computational statistics, Bayesian methods, meta-analysis, and applied decision analysis. He also created PyMC, a library to do probabilistic programming in python, and is the author of several tutorials at PyCon and PyData conferences.