Below are some suggestions to help you get started:
•Using bootstrap algorithm to estimating model coefficients, or population statistics
•Implement a random variable simulation algorithm (such as Metropolis-Hastings)
using different type of proposal functions
•Using non-parametric methods to estimate optimal curve for bi-variate data
Dataset ideas:
and here are some ideas for datasets (click for link)
•Kaggle
•UCI Machine Learning repository
•Pre-loaded datasets in R
R package references:
Below are some example of R packages that could be applied in your final project.
Check for examples using these packages for more ideas:
•R package for kernel density estimation: https://cran.rproject.org/web/packages/ks/vignettes/kde.pdf
•R package for Markov chain Monte Carlo simulation: https://cran.rproject.org/web/packages/mcmc/mcmc.pdf
•R package for bootstrap algorithm: https://cran.rproject.org/web/packages/boot/index.html
Detailed guideline:
While you are not expected to build a computationally complex model, your work needs
to show logical flow, and demonstrates the Bayesian analysis concepts discussed in the
course. This includes the following:
- Description of the problem: What is the problem you are trying to solve? What is the
motivation and significance behind this? Why might your approach be useful here? - Description of your data: What are the variables of interest and their summary? What
are some caveats of the data (such as data quality issues) that we need to be aware of,
if any? - Formulation of your analysis approach: How is the model or estimation algorithm
defined? - Computational approach: What methods are you using to analyze the data? You are
encouraged to use existing R packages. - Results and conclusion: What is the takeaway from your analysis? What makes your
approach advantageous (or challenging) in your problem? What are the next steps in
your analysis?
