Bayesian Statistics for Data Science

Instructor: Dr. Yuxiao Hung | https://www.linkedin.com/in/yuxiao-huang-a00b407a/

COURSE DESCRIPTION:

This course is an introduction of Bayesian data analysis. Topics include parameter estimation (using formal analysis, grid approximation, and Markov chain Monte Carlo), hierarchical models, generalized linear models, JAGS, and Stan. Although lectures will include some theory, the emphasis of the course will be on programming these models in R, JAGS, and Stan, and applying the models to solve real-world problems. This computational aspect will differentiate this course significantly from the Statistics department’s course in Bayesian Methods (STAT 6223).

LEARNING OBJECTIVES:

As a result of completing this course, students will be able to
• Use R, JAGS, and Stan to conduct Bayesian data analysis (parameter estimation, prediction,and model comparison).
• Evaluate Bayesian methods against other approaches on real-world data.
• Write technical report and present the results.
• Work both individually and as a team.

Final Project: Medical Cost Personal Datasets

Insurance Forecast by using Linear Regression and MCMC for Final Project for DATS 6450: Bayesian Statistics by Brent Skoumal (https://github.com/b-skoumal) & Prince Birring (https://github.com/princebirring)

Identify & Acquire

Acceptable performance for this standard is based on how well you've applied specific learning goals within your deliverable. To review the full list of data science standards, see the course syllabus.

1. Well-articulated problem statement with "specific aim" and hypothesis, based on your lightning talk. 2 .An outline of any potential methods and models.
2. Detailed explanation of extant data available.
3. Describe any outstanding questions, assumptions, risks, caveats.
4. Define your goals and criteria, explain what success looks like.
5. Demonstrate domain knowledge, including features or benchmarks from similar projects.

Parse & Mine

Acceptable performance for this standard is based on how well you've applied specific learning goals within your deliverable. To review the full list of data science standards, see the course syllabus.

1. A well organized iPython notebook with code and output.
2. At least one visual for each independent variable (and any relationships) using a python visualization tool.
3. Provide insight about data set and its impact on your hypothesis.

Refine & Build

Acceptable performance for this standard is based on how well you've applied specific learning goals within your deliverable. To review the full list of data science standards, see the course syllabus.

1. Create iPython Notebook with code, visualizations, and markdown.
2. Summarize your exploratory data analysis.
3. Frame source code so it enhances your explanations.
4. Explain your choice of validation and prediction metrics.
5. Include a separate python module with helper functions.
6. Visualize relationships between Y and two strongest variables.
7. Identify areas where new data could help improve the model.

Present & Communicate

1. Include project TOC, background, problem, and hypothesis.
2. Describe datasets and analysis with summary and charts.
3. Demonstrate your model with visualizations.
4. Review the conclusions from your findings.
5. Create a list of recommendations and next steps based on your work.
6. Frame your materials for a non-technical audience.
7. Include an appendix with full technical details.

Prezi Presentation Link:

Link: https://prezi.com/icohm5c8pwce/insurance-forecast-by-using-linear-regression-and-mcmc/