Becoming a BayeZian Bundle | The Complete Bayesian Education | AthlyticZ
Bundle • Save $449
The Complete Bayesian Education
39 modules. 65+ hours. From first principles to shipping physics-constrained GP models in production Stan. One instructor. One price. The full journey.
39 Modules58+ Stan Files65+ HoursCloud VMs Included
Dr. Scott Spencer
Columbia University Professor • Stan Language Collaborator
Columbia•Stan Collaborator•Fortune 500 Advisor
Official Partner
Every student gets access to Posit Workbench — enterprise cloud virtual machines used by pharma, finance, and Fortune 500 data teams. Access provided through AthlyticZ Academy LMS.
The Business Case
Why Your Organization Should Fund This
The cost of one bad analytical hire dwarfs the price of training the right person.
$2,449
This Bundle
39 modules, 65+ hours, production Stan, cloud VMs included
$15K+
University Equivalent
Two semesters of grad Bayesian coursework
$85K+
Cost of a Bad Hire
Recruit, onboard, and replace one failed senior DS
Send This to Your Manager
Request L&D Funding
For less than 2% of my salary, I gain production Bayesian skills that would take 2 years of self-study.
Tax-deductible for employers
Certificate for HR records
Immediate ROI on projects
Cloud VM Access Included
The Journey
From First Principles to Production Stan
Every module builds on the last. Click any module to see all lessons. Follow the wire.
I
BayeZian I • Foundations
Learn to Think in Probability
5 modules • 31 lessons • ~8 hours
Before you write a single line of Stan, you internalize how uncertainty works. This phase rewires your statistical intuition from the ground up.
01
Introducing Bayesian Analysis
Course goals, the Bayesian approach, and why it matters.
2 lessonsFree Preview
Course Introduction
Why Bayesian?
02
Exploring Uncertainty & Variation
Olympic sprint data to build intuition for distributions.
The three pillars. Conjugate priors for efficient computation.
5 lessonsFree Preview
Prior Distributions
Likelihood Functions
Posterior Distributions
Conjugate Priors
Posterior Summaries
05
Simulating Distributions in R
Build intuition through simulation before touching Stan.
4 lessonsFree Preview
Simulating from Known Distributions
Monte Carlo Estimation
Rejection Sampling
Importance Sampling
Why This Matters
Uncertainty Is the Language of Modern Decision-Making
Every serious analytical organization — from the FDA to Goldman Sachs to MLB front offices — requires probabilistic reasoning. Point estimates are not enough. If you cannot quantify uncertainty, you cannot sit at the table where decisions get made.
100%
of top sports teams use Bayesian
$180K+
avg senior Bayesian salary
II
BayeZian I • First Stan Models
Write Your First Bayesian Models in Stan
5 modules • 14 lessons • ~4 hours
Hands on the keyboard. Stan programs, HMC sampling, cmdstanr workflows, grid approximation, and the diagnostics that separate reliable inference from noise.
06
Random Variable Code Objects
Represent distributions as reusable R abstractions.
1 lesson
Random Variable Code Objects in R
07
Simulations and Models in Stan
First Stan programs: beta-binomial, sampling, diagnostics.
5 lessonsFree Preview
Your First Stan Program
Beta-Binomial in Stan
Sampling & Diagnostics
cmdstanr Workflow
Model Comparison Basics
08
Grid Approximation
Discretize the posterior before MCMC.
3 lessonsFree Preview
Grid Approximation Concept
Implementing Grid Approximation
Comparing Grid to MCMC
09
Using the Posterior
Credible intervals, predictions, and decision-making.
3 lessons
Posterior Summaries
Credible Intervals
Posterior Predictions
10
Exchangeability
When data points are similar but not identical.
2 lessons
Exchangeability Concept
Partial Exchangeability
Why This Matters
Stan Is the Gold Standard — and Hiring Managers Know It
Stan powers production models at Amazon, J&J, Pfizer, and every major sports front office. It is explicitly listed in job postings for senior data scientist roles at $120K to $200K+. Learning Stan is not optional if you want to work at the highest levels.
III
BayeZian I • Regression to GLMs
From Regression to Generalized Linear Models
5 modules • 20 lessons • ~6 hours
Normal regression, posterior predictive checks, categorical predictors, the logit link, and binomial outcomes. You become dangerous with real data here.
11
Normal Regression in Stan
Code a full regression model, understand every block.
5 lessonsFree Preview
Simple Linear Regression in Stan
Multiple Regression
Posterior Predictive Distribution
Model Diagnostics
Residual Analysis
12
Posterior Predictive Checks
Does your model generate data that looks like reality?
4 lessonsFree Preview
PPC Concept
Density Overlay Checks
Test Statistic Checks
Calibration Checks
13
Categorical Predictors
Dummy coding, contrasts, and group comparisons.
3 lessonsFree Preview
Dummy Coding
Contrasts
Group Comparisons
14
The Logit Link
From linear to logistic — modeling probabilities.
3 lessonsFree Preview
The Logit Link Function
Logistic Regression in Stan
Interpreting Log-Odds
15
Binomial Outcomes & Soccer Data
GLMs with count data, applied to real soccer analytics.
5 lessonsFree Preview
Binomial GLM
Soccer Shot Data
Expected Goals Basics
Model Comparison
Predictive Performance
Career Milestone
This Is Where You Become Hirable
Regression, GLMs, posterior predictive checks, and categorical predictors are the minimum bar for data scientist roles in sports analytics, pharma, and finance. After this phase, you can build and validate the models that front offices actually use.
$90K-$140K
MLB Senior Quant Analyst
$100K-$150K
NFL Senior Data Scientist
IV
BayeZian I • Hierarchical & Capstone
Think in Levels. Ship a Real Model.
5 modules • 21 lessons • ~7 hours
Hierarchical models, partial pooling, and a full soccer xG case study from raw data to Bayesian decision-making. You finish Part I with a portfolio piece.
16
First Categorical Model
Multi-category outcomes with Bayesian methods.
3 lessonsFree Preview
Multinomial Models
Categorical Predictors
Softmax Regression
17
Hierarchical Models Introduction
Partial pooling: the most important concept in applied Bayesian.
6 lessonsFree Preview
Why Hierarchical?
Partial Pooling
Varying Intercepts
Varying Slopes
Centered vs Non-Centered
Hierarchical Priors
18
Priors for Hierarchical Models
Choosing priors that let the data speak.
3 lessons
Half-Normal Priors
LKJ Correlation Prior
Prior Predictive Checks
19
Case Study: Soccer xG
Full end-to-end project with real soccer data.
7 lessonsFree Preview
xG Data Exploration
Feature Engineering
Building the xG Model
Model Diagnostics
Posterior Predictions
Model Comparison
Final xG Pipeline
20
Bayesian Decision Analysis
Turn posteriors into actionable decisions.
2 lessonsFree Preview
Decision Theory Basics
Loss Functions
The Transition Point
Part I Complete. You Think Like a Bayesian.
You have mastered priors, posteriors, GLMs, hierarchical models, and shipped a real case study. Most data scientists stop here. The ones who keep going become the people who design the models everyone else uses. Part II is where you separate from the pack.
You Are Halfway Through the Journey
The bundle includes everything above and below. One purchase, the full transformation.
Multi-level structures that propagate uncertainty correctly.
4 lessonsFree Preview
Multi-Level Structures
Cross-Classified Models
Nested Random Effects
Uncertainty Propagation
06
Sufficient Statistics
Compress data without information loss for faster inference.
3 lessonsFree Preview
Sufficient Statistics Theory
Implementation in Stan
Performance Gains
Why This Matters
Real Data Is Messy. These Models Handle It.
Overdispersion, zero-inflation, and complex ranking structures show up in every real dataset. Pharma trials have excess zeros. Sports data has overdispersed counts. Player rankings need principled comparison. This phase gives you what basic models cannot.
VI
BayeZian II • Time, Survival & Dynamics
Model What Changes
6 modules • 40 lessons • ~12 hours
Correlation, QR reparameterization, AR processes, survival analysis, and ODE-based dynamics. Time enters the picture.
Survival models are required for FDA drug approval submissions. Time series and AR models drive quantitative finance. ODE-based dynamics are used in aerospace, biomechanics, and sports science. This single phase qualifies you for senior roles in industries most data scientists cannot even interview for.
$180K-$250K
Principal Biostatistician
$200K-$400K+
Staff Quant, Finance
VII
BayeZian II • The Frontier
Nonparametric & Physics-Constrained Inference
4 modules • 46 lessons • ~14 hours
Splines, full Gaussian Processes, Hilbert-space approximations, and physics-constrained likelihoods. This is principal/director territory.
13
Splines & Tensor Products
B-splines, 2D tensor products, Kronecker tricks for xG.
15 lessonsFree Preview
B-Spline Basics
Spline Regression in Stan
Knot Selection
2D Tensor Products
Kronecker Tricks
Penalized Splines
Application: xG Surfaces
Spline Diagnostics
Spline vs GP Comparison
Tensor Product Interactions
Adaptive Knots
Multivariate Splines
Application: Shot Maps
Spline Priors
Spline Summary
14
Gaussian Processes
Full GP priors, Cholesky decomposition, hyperparameter tuning.
10 lessonsFree Preview
GP Theory
Kernel Functions
GP in Stan
Cholesky Implementation
Hyperparameter Tuning
Marginal Likelihood
Multi-Output GPs
Application: Player Curves
GP Diagnostics
GP Summary
15
Hilbert-Space Approximate GPs
HSGP basis functions for scalable nonparametric modeling.
12 lessonsFree Preview
HSGP Theory
Basis Function Approximation
Choosing m and L
1D HSGP in Stan
2D HSGP
Additive HSGP Models
HSGP vs Full GP
Application: Spatial xG
Scalability Benchmarks
Hierarchical HSGP
HSGP Diagnostics
HSGP Summary
16
Physics-Constrained Models
Sail GP, golf putting, base running, umpire strike zones.
9 lessonsFree Preview
Physics-Informed Priors
Sail GP: Wind & Speed
Golf Putting: Geometry
Base Running: Kinematics
Strike Zone: Boundaries
Multi-Physics Models
Physics Model Diagnostics
Constrained Parameters
Physics Summary
The Differentiator
Fewer Than 1% of Data Scientists Can Build These Models
Gaussian Processes, HSGP approximations, and physics-constrained likelihoods separate a $120K data scientist from a $250K+ principal. Amazon uses hierarchical GPs for personalization. SpaceX uses physics-constrained Bayesian models for launch reliability.
$220K-$350K
Staff DS, Big Tech
$175K-$275K
Director of Quant, Sports
VIII
BayeZian II • Ship It
Production Performance & What Comes Next
3 modules • 16 lessons • ~5 hours
Robustness, missing data, censoring, GPU acceleration, parallelization, and memory management. Make your Stan models production-grade.
17
Common Issues
Outliers, missing data imputation, censoring, truncation.
Your roadmap: causal inference, state-space, advanced GPs.
1 lesson
What Comes Next
Journey Complete
You Do Not Just Know Bayesian. You Ship It.
GPU-accelerated Stan, parallelized likelihoods, robust handling of messy data — these are the production engineering skills that turn a great analyst into someone who builds and deploys models that organizations actually run on.
Columbia University Professor • Stan Language Collaborator
Dr. Scott Spencer is one of the world's foremost experts in applied Bayesian analysis. As a Columbia professor and Stan collaborator, his methods power decisions at Amazon, Johnson & Johnson, Vevo, and leading sports franchises. This bundle represents the complete transfer of Scott's applied Bayesian knowledge — from the foundations he teaches at Columbia to the production patterns he uses in Fortune 500 consulting.
Columbia University
Stan Collaborator
Fortune 500 Advisor
The Complete Bayesian Education
From First Principles to Production Stan
39 modules. 65+ hours. 58+ Stan files. Cloud VMs included. Two certificates. One price.
Official Partner