Statistical Rethinking With Brms Ggplot2 And The Tidyverse Versions Save

Noteworthy changes include:

• replacing my incorrect use of tidyr::expand() with a more appropriate tidyr::expand_grid() workflow,
• adopting the new linewidth argument for several ggplot2 geoms, and
• refitting all models with the current official version of brms (2.18.0).

Noteworthy changes include:

• the correct solution to the first multinomial model in Section 10.16;
• a coherent workflow for the Gaussian process model from Section 13.4;
• corrections to some of the post-processing workflows for the measurement-error models in Section 14.1;
• refitting all models with the current official version of brms, version 2.13.5;
• improved in-text citations and reference sections using BibTex, Better BibTeX, and zotero; and
• minor prose, hyperlink, and code edits throughout.

Noteworthy changes include:

• substantial expansions to sections 10.3.1 (multinomial regression) and 11.3.2 (negative binomial regression),
• the addition of a new section in Chapter 15 (15.9) encouraging others to code in public,
• refitting all models with the current official version of brms, version 2.12.0,
• discussions (8.3.2) on the new Bulk_ESS and Tail_ESS summaries of HMC effective sample size (Vehtari et al., 2019),
• saving all fits as external files in the new GitHub fits folder, primarily with the file argument,
• extensive use of the patchwork package for combining ggplots,
• improving/updating some of the tidyverse code (e.g., using tidyr::crossing()),
• updates to the brms::custom_family()-related code in 11.3.1 to match brms 2.11.0 updates,
• replacing the depreciated brms::marginal_effects() with brms::conditional_effects() (see issue #735),
• replacing the depreciated brms::stanplot() with brms::mcmc_plot(),
• increased the plot resolution with fig.retina = 2.5,
• refreshed hyperlinks

Along the way to making the PDF version of the project, we

• Updated the LaTeX code (thanks to Stijn Masschelein)
• Updated a few hyperlinks
• Reformatted a few code blocks
• Updated the Bayesian $R^2$ reference (thanks to Aki Vehtari)