Conclusions
Bayesian models are good
- Increasing levels of assumptions: Least Squares, to Maximum Likelihood-based models, to Bayesian models.
- We’re standing on the shoulders of giants! Be mindful and use prior knowledge.
Different flavors of MCMC have enabled applied Bayesian models
- MCMC has enabled modern applied Bayesian models.
- Computational efficiency of MCMC is improving.
Don’t forget to check
- No free lunches: powerful computations still need babysitting.
- Model checking helps understanding your model’s strengths and weaknesses.