First meeting (HMC): Summary & Questions

Summary

• Began to know about the workflow when using R as suggested. https://www.tellingstorieswithdata.com/workflow.html#developing-research-questions. Basic knowledge: R Markdown, R projects, relevant skills to access data, write code comments, etc.
• Set up R project for all the documents used to practice with: Example data, R scripts, RMarkdown files and output files, Summary documents for each meeting, etc.
• Have installed rstan and brms packages in RStudio.
• Read Chapter 1,2,3 and 9 of “Statistical Rethinking” using rstan as well as brms: https://bookdown.org/content/4857/markov-chain-monte-carlo.html. Reproduced the output of the “Terrain Ruggedness” example: “rugged” data pre-processing, sampling from posterior probability of 3 parameters, how to set up # of chains & cores, sample visualizations, check the chains with trace plots and trank plots, tuning superparameters like iter, warmup.
• Had a glance at the recommended article (comparison between BUGS and STAN): https://besjournals.onlinelibrary.wiley.com/doi/epdf/10.1111/2041-210X.12681.
• Materials about origin or concepts of HMC that I read once or am currently reading:
  • Neal’s review, MCMC using Hamiltonian dynamics (Neal, 2012): https://arxiv.org/pdf/1206.1901.pdf
  • Tutorial notes (Manchester) about MCMC & HMC.
  • Hamiltonian Monte Carlo for Hierarchical Models (Mark, 2013): https://www.researchgate.net/publication/259106135_Hamiltonian_Monte_Carlo_for_Hierarchical_Models
  • A Conceptual Introduction to Hamiltonian Monte Carlo (Michael, 2018): https://arxiv.org/pdf/1701.02434.pdf
• BRMS package preferred: Now reading “Advanced Bayesian Multilevel Modeling with the R Package brms” (Paul-Christian Bürkner, 2018): https://journal.r-project.org/archive/2018/RJ-2018-017/RJ-2018-017.pdf. Trying all the examples included in this article.
• Tried to use RMarkdown to present the output, analysis results of some examples for practice. (saved as .html). Learning some commands commonly used.

Questions

• Requirement for numbers of words or pages?
• Presents novel models/methods?
• Medical data, where to get?
• A timeline of the statistic project.
• Next steps:
  • continue exploring brms.
  • start looking for the topics and data for the final dissertation.
  • review and learn more about HMC.
  • use RMarkdown to write the general framework for the dissertation.
  • Q: What else?
• Leapfrog method. Neal’s review, MCMC using Hamiltonian dynamics (Neal, 2012): https://arxiv.org/pdf/1206.1901.pdf, page 8 & 14

   knitr::include_graphics("images/leapfrog algorithm.png")

   knitr::include_graphics("images/leapfrog raw codes.png")