Bayesian Data Analysis Final Project: Fall 2023
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Bayesian Data Analysis Final Project: Fall 2023
1 Project
For the final project for the course, your assignment is essentially to perform a complete Bayesian data analysis on data of your choice. The final project is worth 15% of the final grade, so think of it as the equivalent of about two weeks of course work.
2 Procedure and Dates
. Your entire submission is due at noon, Tuesday December 19, 2023.
. This is an individual, not a group, project.
. Submit your project via Canvas. The submission consists of four (in some cases, three) things:
1. An R Markdown file ( .Rmd).
2. *Any necessary data file(s) that you used (see below).
3. A Notebook file ( .nb.html). The idea is that the TAs and I should be able to execute your .Rmd file, with the submitted data file(s) placed in the same folder, and completely reproduce this .html file. (Note: if you’re sourcing the data file directly from some web address, then there is no need to submit the data file, so long as the .Rmd file can reproduce the .nb.html file.)
4. A one-minute video (minimum 55 seconds, maximum 65 seconds) in which you present your project. You must appear in the video the entire time as you speak.
. Do not use a dataset that you have used in this or any other course.
. I expect that you will discuss this project with others, but please avoid using datasets in common (I realize that might still happen by coincidence). All of the work sub- mitted must be your own. Be sure to credit the sources of your data and any other material – it is better to over-credit than to under-credit. If you have any questions about properly crediting others’ work, you should contact me about it.
. The last lines of your R Markdown file should be “ On my honor, I have neither received nor given any unauthorized assistance on this project,” followed by your name. The project is not anonymous.
3 Required Elements
. Make clear where the dataset comes from, and include a description of the dataset.
. Include more than one predictor variable.
. Include trace plots, R(ˆ) values, and effective sample sizes to demonstrate that the
MCMC process has worked well.
. Summarize posteriors, both numerically and graphically.
. Include a posterior predictive check.
. Use your model to make predictions at some particular set of predictor values.
. Include some interpretation of your results.
. Since the TAs and I will be reading your code, your code should be easy to understand
– use spaces, don’t be shy about using new lines, and don’t use names like x when you could use variable names like income.
. All of your project should be written up in an R Markdown (i.e., .Rmd) file in which you mix text (outside of the code chunks) and code (inside the code chunks).
In addition, for your one-minute video presentation, you’re expected to give a brief and complete presentation on all aspects of your analysis. You will be assessed based on the comprehensiveness, coherence, and precision of your presentation. Adhere to the time limit. (Just to give you an idea, a typical person can speak about 140 English words per minute, give or take about 30 words.) You may choose to screen-share during the video (this is not required). However, whether you screen-share or not, you must appear in the video the entire time as you speak.
Roughly speaking, your presentation video constitutes about 30% of your project grade. However, the grading of the project will be done holistically, taking all aspects into account.
4 Additional Elements
The following will elements are not required, but will favorably affect your grade. An A on the project is not possible if none of these elements are included.
. A binary or count variable as the outcome.
. A hierarchical model.
. A model outside those discussed in the course.
. Use of priors other than the defaults (along with an explanation of your choices).
. Comparison among more than one model.
. The use of loo as a diagnostic.
. A prior predictive check.
. Some statistical inference based on the posterior (e.g., the probability that a regres- sion coefficient is greater than zero, and the practical significance of that).
. Other aspects that go beyond a vanilla analysis.
5 A Final Note
You will both enjoy the project more and you will do better if you pick something that you are genuinely interested in. You may also find yourself expanding the project somewhat so that you can brag about it in a job interview – it’s hard to make someone interested in a project that you are not interested in yourself.
2023-12-20