About
1
An introduction to Stan
1.1
Why might you want to start learning Bayesian methods?
Benefit 1: Incorporating knowledge from outside the data
Benefit 2: Combining sources of information
Benefit 3: Dealing with uncertainty consistently in model predictions
Benefit 4: Regularizing richly parameterized models
Benefit 5: Doing away with tests
1.2
Models and inference
1.3
Why use Stan?
1.4
Background: Bayes rule, likelihood and priors
1.4.1
Likelihood/log likelihood
1.4.2
So what does the likelhood mean?
1.4.3
Prior distributions
1.4.4
Bayes rule
1.5
HMC and Betancourt
1.6
A tour of a Stan program
1.6.1
A hello world example
2
Modern Statistical Workflow
2.1
Modern Statistical Workflow
2.1.1
Example: A simple time-series model of loan repayments
2.1.2
Step 1: Writing out the probability model
2.1.3
Step 2: Simulating the model with known parameters
2.1.4
Model inspection
2.1.5
Taking the model to real data
2.2
Tools of the trade: borrowing from software engineering
3
A more difficult model
4
Aggregate random coefficients logit: Bayesian estimation using Stan
4.1
A generative model of consumer choice
4.1.1
Generating aggregate sales data from the model
4.1.2
Modeling price
4.1.3
Estimating the model from aggregate market-level data
4.2
Part 2: Fake data simulation
4.3
Part 3: Writing out the model in Stan
4.4
Conclusion
A brief introduction to econometrics in Stan
Session 3
A more difficult model