Bayesian-Computation-Project

Research

• Engaged in a comprehensive project aiming to master Bayesian methods for regression models, especially when dealing with binary or polytomous outcome variables, or multiple outcome variables in R studio
• Analyzed heart-attack dataset and applied advanced techniques, Cumulative Link Model, Hamiltonian Monte Carlo, Horseshoe Prior to establishing Bayesian model and making prediction

Conclusion

• Bayesian model with horseshoe prior performs the best
• 1 (male) is associated with the log odds decrease in the outcome by 1.64 - Female is more likely to get heart attack. Typical chest pain would have a higher impact on the outcome variable compared to other type of chest pains.

Members

• Haiyue Zhang, Wendy Zheng, Ishaan Bhandari, Kehuan Wang, April Wu

Mentor

• Professor Steven A. Culpepper