Supplementary material code for “Extended-support beta regression for [0, 1] responses”

Ioannis Kosmidis, Achim Zeileis

March 17, 2025

Directory structure

Warning: package 'topmodels' was built under R version 4.4.3

The directory code/ contains the scripts

script
beta-vs-4-par-beta.R
beta01.R
lossaversion.R
reading-skills-interaction-stats.R
xbx-mean-variance.R
xbx-vs-beta-crps.R
xbx-vs-beta.R
xbx-vs-cn-crps.R
xbx-vs-cn.R
xbx-vs-ols-crps.R

that reproduce all the numerical results and figures in the manuscript

Kosmidis, I. and Zeileis, A. (2024). Extended-support beta regression for [0, 1] responses. https://arxiv.org/abs/2409.07233

and the supplementary material document xbx-supp.pdf.

The directory results/ is populated by R image files that store the numerical results the scripts produce. The directory figures/ is populated by graphics that the scripts produce.

R version and contributed packages

All results are reproducible using R version 4.4.2 (2024-10-31) and the contributed packages

package	version
betareg	3.2-1
crch	1.2-1
distributions3	0.2.2
dplyr	1.1.4
ggplot2	3.5.1
lmtest	0.9-40
topmodels	0.3-0
VGAM	1.1-12

At the time of writing the topmodels package is not on CRAN. Please install it by doing

install.packages("topmodels", repos = "https://zeileis.r-universe.dev")

Reproducing the results

The results can be reproduced by using an interactive R session and going through the following scripts line-by-line:

1. code/reading-skills-interaction-stats.R reproduces

• Figure 1 in the main text;

2. code/beta-vs-4-par-beta.R reproduces

• Figure 2 in the main text;

3. code/xbx-vs-beta.R reproduces

• Figure 3 in the main text;

4. code/xbx-vs-cn.R reproduces

• Figures S1-S2 in Section S2 of the supplementary material document;

5. code/xbx-mean-variance.R reproduces

• Figures S3-S4 in Section S3 of the supplementary material document;

5. code/lossaversion.R reproduces

• the numerical figures in Tables 1-3 of the main text,
• Figures 4-5 of the main text,
• the Wald and likelihood ratio statistics reported in Section 4.5 of the main text,
• Figure S5 in Section S4 of the supplementary material document,
• the numerical figures in Tables S1-S3 in Section S4 of the supplementary material document;

6. code/xbx-vs-cn-crps.R carries out the numerical experiment in Section 5, and reproduces

• Figure 6 of the main text,
• Figures S6-S10 in Section S5 of the supplementary material document.

Computation here relies on parallel computing, which is implemented through the parallel R package. The script will not work on Windows unless n_cores <- 1 (which will lead in long compute times and is not recommended) or it is modified to use a different parallel back-end. All results should be exactly reproducible in Unix-based systems (e.g. macOS and Linux).

7. code/xbx-vs-cn-beta.R and code/xbx-vs-cn-ols.R carry out the numerical experiments in Section S6 of the supplementary material document, and reproduce

• Figures S11-S12 of the supplementary material document