saros: Semi-Automatic Reporting of Ordinary Surveys

Introduction

{saros} (Semi-Automatic Reporting of Ordinary Surveys) is an R package designed to handle repeating surveys within the same project that occur annually or biannually. It aims to automate the process of summarizing and reporting on survey data, helping researchers save time and maintain consistency across survey iterations. Specifically, {saros} produces highly customizable figures, tables, analyses and complex interactive reports for a batch of possible dependent-independent relations of possible interest.

Overview: tools for five stages of the report production

What	Note	Sub-package
Project setup containing ready-made R-scripts and folders	Optional	saros.base
Data cleaning	Only supplements {tidyverse}/{datawizard}/ {labelled}-tools	saros.utils (not on CRAN)
Report drafting		saros.base
Easy content generation for common standardized outputs	Standardized output types. Alternatively use your own functions	saros
Web access restriction and distribution	Optional	saros.base

Why saros?

• Simplicity: Setting up a reporting system, in particular a semi-automated one in Quarto, can be daunting - in particular if some of the chapter authors/collaborators have little familiarity with R/Python or Quarto/RMarkdown.
  • Created figures are minimal, meaning that you are given full power to adjust them post-hoc using usual ggplot2::theme() tools, including ggplot2::set_theme().
• Flexibility:
  • Several sets of chunk templates are built in, depending on the type of report you want. These templates are provided as data frames, so you can easily adjust these or create your own.
  • A multitude of settings with optional glue-templating of prefixes, infixes and suffixes allows translations and adaptations.
• Consistency and reproducibility: Ensure all your outputs within a chapter, a report, a project, or even an organization, use the same formatting and structure. Figures and plots are purposely simple in theming and aesthetics, so that the end-users can specify this themselves:
  • When editing/rendering reports, use global options (and override when necessary). For the drafted report chunks, the saros content functions for plots, tables, etc use inheritance for finding its settings:
    • If specified in the function, it will use that setting.
    • If not specified, it will check global options, which can be specified for the chapter (qmd-file), the entire report, the entire project, or the organization’s settings.
    • If none of the above is specified, function factory defaults are used.
  • Also link creation to automatically created
  • Convenience functions set, get and reset options
  • Instead of using probabilistic AI-tools, ensure your reports always come out as expected.
• Aesthetics and accessibility:
  • Opinionated, yet highly flexible, interactive {ggiraph}-based figures (building on ggplot2) and gt-compatible tables.
• Performance: The saros tools draft a report in 3-4 seconds. Spend the remaining time thinking about what to write.
  • Although built-in plotting functions use (the somewhat slower performing) ggplot2, it is easier for the majority to modify such plot objects, and to expand with their own compliant plotting functions. One can also easily insert other base/lattice-plotting functions in the report-drafting templates.

Workflow

Preliminaries:

1. Optionally set up your project directory for a completely new project.
2. Clean your raw data: - Variables should be stored in the data type that they should be displayed as (factor, ordered factor, integer, character, etc). Ordered factors will in certain outputs be kept in the given order, whereas a set of unordered factors may be e.g. sorted_by = ".upper" (e.g values of the upper-most categories). - Variables should have variable labels, and sets of variables should have the same variable label prefix. Prefix and suffix can be split by e.g ” - “. Use e.g. {labelled} for most operations. For advanced cleaning, see saros.utils

{saros}-tools

• makeme() makes most types of output for your report you would need for surveys. The function can be extended with S3-methods tailored for your needs.
  • The core idea behind this function is the reusability of global options, which makes it possible to globally adjust all outputs with a small switch.
• make_link() will upon rendering save a plot, dataset or any other object to disk and return a “download plot”-link with a unique (hashed) filename.
• n_range() returns the sample size range given a dataset, dependent variables and independent variables.
  • An alternative n_range2() allows directly using a makeme() output.
• fig_height_h_barchart() estimates the best figure height for a horizontal barchart, based on a data frame, dep and indep variables, and other arguments.
  • An alternative fig_height_h_barchart2() takes a makeme(type="cat_plot").

{saros.base}-tools

1. Optionally specify chunk templates for what you want for each set of related variables. Or use among the built-in sets of templates.
2. Optionally, create a chapter_overview (a compact description of which dependent (and independent) variables goes in which chapter-file). A data.frame where a row is a chapter. Must contain at least the columns ‘chapter’ and ‘dep’. ‘dep’ uses tidyselect-syntax in each cell. If not using any, all variables are processed and placed in the same qmd-file. More details
3. Combine the raw data, chunk templates and chapter_overview to make a chapter structure. This will create a data frame containing your report structure, which can be further tailored. Arguments to refine_chapter_overview() allows e.g ignoring:
   • non-significant bivariate relationships between dependent and independent variables,
   • low observation counts (sample size) for categories, variables or dependent-independent cell combinations, and/or
   • variables with all NA for a given “crowd” (a target group, all others combined, or all)
4. Draft the report, using the output from refine_chapter_overview() and your raw data. Optionally provide a range of YAML-defaults and QMD-prefixes/suffixes to your chapter-files, index-files or full report-files, as well as creating chapter-datasets for compartmentalized authoring.
5. After rendering your Quarto Project (using regular Quarto/RStudio tools), optionally batch configure access restrictions and send out glue-tailored emails to institutions that have participated in your survey, now receiving password-protected access to their own report.
6. Having done this once for a report, you might want to create a project template for your organization. Folder structures can be mapped to a YAML file so that it can later be easily created at once with your preferred directory numbering scheme. If located on e.g. GitHub this can be downloaded, unzipped and placed in a new project folder.

Installation

The development version from GitHub with:

install.packages("pak")
pak::pak("saros")
# # pak::pak("NIFU-NO/saros")
install.packages("saros", repos = c('https://nifu-no.r-universe.dev', 'https://cloud.r-project.org')) # Alternatively, latest developer-version at GitHub

Draft a simple report using defaults, in a temporary folder

library(saros)
refine_chapter_overview(data = ex_survey,
                        chapter_overview = ex_survey_ch_overview) |>
  draft_report(data = ex_survey)

Draft a customized report

• Dropping

library(saros)
refine_chapter_overview(data = ex_survey,
                        chapter_overview = ex_survey_ch_overview,
                        chunk_templates = get_chunk_template_defaults(2),
                        always_show_bi_for_indep = "x1_sex",
                        hide_bi_entry_if_sig_above = .05) |>
  draft_report(data = ex_survey,
               prefix_heading_for_group = c(".variable_name_indep" = "---\n"),
               serialized_format = "qs" # qs-format, if installed, is faster than rds
               )

Draft a customized report, one for each participating university

• The auxiliary variable f_uni is included in all chapter datasets so that the YAML-header in the QMD-file can refer to it, and chunks later refer to that global parameter:

---
params:
    mesos_var: "f_uni"
    mesos_group: "Uni of A" # This can be replaced in a command: `quarto render ch1.qmd -P mesos_group:'Uni of A'`
---

library(saros)
refine_chapter_overview(data = ex_survey,
                        chapter_overview = ex_survey_ch_overview,
                        chunk_templates = get_chunk_template_defaults(2),
                        always_show_bi_for_indep = "x1_sex",
                        hide_bi_entry_if_sig_above = .05) |>
  draft_report(data = ex_survey,
               prefix_heading_for_group = c(".variable_name_indep" = "---\n"),
               auxiliary_variables = "f_uni",
               serialized_format = "qs"
               )

IMPORTANT: saros is free, except in Norway

Outside of Norway, this package can be used in accordance with the MIT license. However, persons affiliated with Norwegian non-profit or profit/commercial organizations must have the explicit written permission from the author for use. Simply put, if your department/organization competes with NIFU on commissioned or open research calls within Norway, you are probably not going to get permission unfortunately.

Acknowledgements

• The internal code for reordering unordered factors in saros:::make_contents.cat_plot_html() has been copied (without permission) from Julia Silge’s and David Robinson’s excellent {tidytext}-package.
• The code for loading all saros-packages and resolving conflicts has been copied (without permission) from the tidyverse-package and adapted.

Code of Conduct

Please note that the saros project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.