R Markdown Exercises: 25 Real-World Practice Problems

ex_1_1 <- '---
title: "Quarterly KPI Review"
output:
  html_document:
    toc: true
    toc_float: true
    code_folding: hide
---'
cat(ex_1_1)
#> ---
#> title: "Quarterly KPI Review"
#> output:
#>   html_document:
#>     toc: true
#>     toc_float: true
#>     code_folding: hide
#> ---

Explanation: YAML in an Rmd is whitespace-sensitive: two-space indentation under html_document: is the contract. toc_float: true requires toc: true to be set, otherwise it silently does nothing. code_folding: hide collapses chunks by default; use show if you want them open and let readers hide. A common mistake is using tabs instead of spaces, which throws a parser error at render time.

ex_1_2 <- 'output:
  pdf_document:
    latex_engine: xelatex
    fontsize: 11pt
    geometry: margin=1in
    mainfont: "TeX Gyre Termes"'
cat(ex_1_2)
#> output:
#>   pdf_document:
#>     latex_engine: xelatex
#>     fontsize: 11pt
#>     geometry: margin=1in
#>     mainfont: "TeX Gyre Termes"

Explanation: mainfont is only honoured under xelatex or lualatex; the default pdflatex engine ignores it because it cannot load system fonts. The geometry field accepts the LaTeX geometry-package syntax, so you can specify per-side margins like top=1in, bottom=0.75in for asymmetric layouts. Double-quote font names that contain spaces, otherwise YAML treats the trailing word as a separate value.

ex_1_3 <- 'output:
  html_document:
    toc_float: true
  word_document:
    reference_docx: corp_template.docx
  pdf_document:
    latex_engine: xelatex'
cat(ex_1_3)
#> output:
#>   html_document:
#>     toc_float: true
#>   word_document:
#>     reference_docx: corp_template.docx
#>   pdf_document:
#>     latex_engine: xelatex

Explanation: When output: has multiple children, rmarkdown::render() defaults to the first listed format; pick the one your stakeholders see most often. To render every format in one go from the command line, call rmarkdown::render("report.Rmd", output_format = "all"). The reference_docx path is resolved relative to the Rmd file, not the working directory, so colocate the template with the source.

ex_1_4 <- 'bibliography: refs.bib
csl: csl/apa-7th.csl
link-citations: true'
cat(ex_1_4)
#> bibliography: refs.bib
#> csl: csl/apa-7th.csl
#> link-citations: true

Explanation: Pandoc handles citations transparently when these three fields are set; you reference entries inline with [@smith2020] and pandoc replaces them at render time. link-citations: true turns each citation into a hyperlink that jumps to the references section, which readers expect in HTML output. The csl field accepts paths relative to the Rmd; if omitted, pandoc uses its default Chicago-style.

ex_1_5 <- 'params:
  region:
    value: "EMEA"
  cutoff_date:
    value: !r Sys.Date()'
cat(ex_1_5)
#> params:
#>   region:
#>     value: "EMEA"
#>   cutoff_date:
#>     value: !r Sys.Date()

Explanation: The !r tag is a YAML directive that knitr evaluates as R code at render time, which is how you get a live Sys.Date() default rather than a hard-coded string. You can also add input: select with a choices: list to expose a dropdown when rendering interactively. Inside the body, reference these as params$region and params$cutoff_date; they arrive as the types you declared.

ex_2_1 <- "{r exec-summary, echo=FALSE, message=FALSE, warning=FALSE}"
cat(ex_2_1)
#> {r exec-summary, echo=FALSE, message=FALSE, warning=FALSE}

Explanation: echo=FALSE hides the code, results='asis' would mean something different (it tells knitr the chunk output is already markdown). Suppressing messages and warnings is standard for an executive-facing chunk where startup chatter from packages would clutter the page. A common mistake is using include=FALSE, which hides BOTH code and output; that is what you want for setup chunks, not summary ones.

ex_2_2 <- 'knitr::opts_chunk$set(
  message = FALSE,
  warning = FALSE,
  fig.width = 7,
  fig.height = 4,
  fig.align = "center"
)'
cat(ex_2_2)
#> knitr::opts_chunk$set(
#>   message = FALSE,
#>   warning = FALSE,
#>   fig.width = 7,
#>   fig.height = 4,
#>   fig.align = "center"
#> )

Explanation: opts_chunk$set() writes into the knitr session's chunk option defaults, so every subsequent chunk inherits them unless it overrides individual options in its header. Place this call in a chunk with include=FALSE so the defaults change without leaking into the rendered output. Per-chunk overrides always beat the global defaults, which is exactly the pattern you want for one-off "show this code" exceptions.

ex_2_3 <- '{r fit-model, cache=TRUE, cache.extra=file.info("sales.csv")$mtime}'
cat(ex_2_3)
#> {r fit-model, cache=TRUE, cache.extra=file.info("sales.csv")$mtime}

Explanation: cache=TRUE alone hashes the chunk's source code, so the cache stays valid even when the underlying data changes, which is the most painful bug in cached reports. Pinning cache.extra to the file's modification time forces knitr to recompute when the file is touched. You can also use tools::md5sum("sales.csv") to invalidate on content rather than timestamp; pick mtime for cheap files and md5 for ones a build process might re-write with identical content.

ex_2_4 <- '{r mpg-wt, fig.width=6, fig.height=3.5, dpi=150, fig.cap="Fuel economy declines with vehicle weight", fig.alt="Scatterplot of mpg versus weight"}'
cat(ex_2_4)
#> {r mpg-wt, fig.width=6, fig.height=3.5, dpi=150, fig.cap="Fuel economy declines with vehicle weight", fig.alt="Scatterplot of mpg versus weight"}

Explanation: fig.width and fig.height are in inches and feed into the device sizing; dpi controls the raster resolution for HTML and Word output. fig.alt is the accessibility text screen readers announce and is distinct from fig.cap, which appears visibly under the figure. PDF output ignores dpi for vector graphics, so a high dpi only matters for the html and docx pipelines.

ex_2_5 <- "{r setup, include=FALSE}"
cat(ex_2_5)
#> {r setup, include=FALSE}

Explanation: include=FALSE is the only flag you need: it implicitly sets echo=FALSE, suppresses results, hides messages and warnings, and skips figure output. By convention this chunk is labelled setup so RStudio knows to evaluate it first when you run any later chunk interactively. Avoid putting eval=FALSE here by mistake; that would skip the side effects entirely and your later chunks would not see the loaded packages.

ex_3_1 <- "The dataset contains `r nrow(mtcars)` observations."
cat(ex_3_1)
#> The dataset contains `r nrow(mtcars)` observations.

Explanation: Inline R is the single most-used feature in non-trivial reports because it keeps text and numbers in sync; rewrite the sentence, not the source data. The leading r after the opening backtick is what tells knitr to evaluate (vs. styling code as monospace). For values that need formatting, wrap them: ` r format(nrow(mtcars), big.mark = ",") produces 32 here but 1,234,567` on a bigger frame.

ex_3_2 <- knitr::kable(head(mtcars), caption = "Top of mtcars", align = "r")
ex_3_2
#> Table: Top of mtcars
#> 
#> |                  |  mpg| cyl| disp|  hp| drat|    wt|  qsec| vs| am| gear| carb|
#> |:-----------------|----:|---:|----:|---:|----:|-----:|-----:|--:|--:|----:|----:|
#> |Mazda RX4         | 21.0|   6|  160| 110| 3.90| 2.620| 16.46|  0|  1|    4|    4|
#> ...

Explanation: kable() returns a markdown-formatted string when output format is HTML or PDF and pipes through pandoc cleanly; pass format = "html" only if you need raw HTML for further styling. The align argument accepts either a single character like "r" (right-align everything except the row labels) or a per-column string like "lrrrrr". For numeric formatting use the digits argument, which avoids fragile global options(scipen = ...) hacks.

ex_3_3 <- knitr::kable(head(mtcars), format = "html") |>
  kableExtra::kable_styling(bootstrap_options = "striped") |>
  kableExtra::add_header_above(c(" " = 1, "Engine" = 4, "Performance" = 7))
ex_3_3
#> <table class="table table-striped">
#>   <thead>
#>     <tr><th></th><th colspan="4">Engine</th><th colspan="7">Performance</th></tr>
#>     ...
#>   </thead>
#>   <tbody>...</tbody>
#> </table>

Explanation: format = "html" is required because kable_styling() only operates on HTML kables; the default markdown output would silently ignore the styling pipe. The named vector passed to add_header_above() must sum to the total column count including the row-label column, so the leading " " = 1 is mandatory. For PDF output, swap to bootstrap_options = NULL and use latex_options = "striped" instead.

ex_3_4 <- knitr::kable(
  head(mtcars),
  digits = c(2, 0, 0, 0, 2, 2, 2, 0, 0, 0, 0),
  format.args = list(big.mark = ",")
)
ex_3_4
#> |                  |   mpg| cyl| disp|  hp| drat|   wt|  qsec| vs| am| gear| carb|
#> |:-----------------|-----:|---:|----:|---:|----:|----:|-----:|--:|--:|----:|----:|
#> |Mazda RX4         | 21.00|   6|  160| 110| 3.90| 2.62| 16.46|  0|  1|    4|    4|
#> ...

Explanation: Pass digits as a vector of length equal to the data column count to control per-column rounding; pass a scalar for uniform rounding. format.args is forwarded to base R's format() and accepts big.mark, decimal.mark, scientific, and so on. For percentages, format the value upstream with scales::percent() and then drop the column into kable as a character; otherwise digits will fight with your formatter.

ex_3_5 <- "{r diag-plots, fig.show='hold', out.width='50%', fig.ncol=2}"
cat(ex_3_5)
#> {r diag-plots, fig.show='hold', out.width='50%', fig.ncol=2}

Explanation: fig.show='hold' tells knitr to defer rendering plot output until the end of the chunk so multiple plots can be laid out together rather than appearing inline with the print statements that produced them. out.width='50%' is the display width in the final document, distinct from fig.width, which is the device-side rendering width. fig.ncol controls the column count of the resulting grid. For three plots at one-third width, switch to out.width='33%', fig.ncol=3.

ex_4_1 <- sales |>
  filter(region == params$region, sale_dt >= params$cutoff_date)
ex_4_1
#> # A tibble: 1 x 3
#>   region sale_dt    revenue
#>   <chr>  <date>       <dbl>
#> 1 EMEA   2026-03-02     340

Explanation: Inside an Rmd, params is a regular named list available everywhere; you reference fields with $ exactly like any list. The YAML value: !r Sys.Date() ensures cutoff_date arrives as a Date, so date-comparison works without coercion. If you ever see a date comparison silently match zero rows, check class(params$cutoff_date): a string default coerces silently to character and produces lexicographic comparisons that look like dates but aren't.

regions <- c("EMEA", "APAC", "NA")
ex_4_2 <- vapply(
  regions,
  function(r) rmarkdown::render(
    "report.Rmd",
    output_file = paste0("report_", r, ".html"),
    params      = list(region = r),
    envir       = new.env()
  ),
  character(1)
)
ex_4_2
#> [1] "report_EMEA.html" "report_APAC.html" "report_NA.html"

Explanation: Each render() call must use a fresh environment via envir = new.env(), otherwise the second render would inherit the first one's params and side effects, producing wrong-region output that silently looks right. vapply() with character(1) enforces a single-string return type, which catches a render error early rather than letting it surface as a list with an NA element. For dozens of regions, swap to furrr::future_map() to render in parallel processes.

ex_4_3 <- knitr::knit_child(
  "_disclaimer.Rmd",
  envir = environment(),
  quiet = TRUE
)
cat(ex_4_3)
#> ## Disclaimer
#> 
#> This report contains forward-looking statements...

Explanation: knit_child() returns the rendered markdown as a single string, so you wrap the call in an inline R expression: ` r knitr::knit_child("_disclaimer.Rmd") inside body text injects it at that exact spot. The leading underscore on the filename is a convention indicating "not a standalone output", and rmarkdown will skip it when you call rmarkdown::render_site(). Passing envir = environment() lets the child see the parent's variables, including params`.

ex_4_4 <- rmarkdown::render(
  "report.Rmd",
  params = list(region = Sys.getenv("REPORT_REGION", unset = "EMEA"))
)
ex_4_4
#> [1] "report.html"

Explanation: Sys.getenv() with unset = "EMEA" provides a fallback when the env var is missing; without it, an unset variable returns the empty string, which silently produces a report filtered to zero rows. CI pipelines normally export the variable before invoking R: REPORT_REGION=APAC Rscript -e 'rmarkdown::render(...)'. For secret values (API tokens) prefer Sys.getenv() over passing them on the command line, where they end up in shell history.

library(purrr)
library(tidyr)
grid <- expand_grid(region = c("EMEA","APAC"), product = c("widgets","gadgets"))
ex_4_5 <- pmap_chr(grid, function(region, product) {
  rmarkdown::render(
    "report.Rmd",
    output_file = paste0("report_", region, "_", product, ".html"),
    params      = list(region = region, product = product),
    envir       = new.env()
  )
})
ex_4_5
#> [1] "report_EMEA_widgets.html" "report_EMEA_gadgets.html"
#> [3] "report_APAC_widgets.html" "report_APAC_gadgets.html"

Explanation: pmap_chr() runs the function once per row of grid and binds the results into a character vector, type-checked to be exactly one string per row. expand_grid() is the tidyverse equivalent of base R's expand.grid() but preserves column order and returns a tibble. For a sparse subset (only EMEA-widgets and APAC-gadgets), build the tibble directly with tribble() rather than a full grid then filter, which keeps intent visible to the reader.

ex_5_1 <- "The drift detection approach builds on prior work [@smith2020, p. 14].\n\n## References"
cat(ex_5_1)
#> The drift detection approach builds on prior work [@smith2020, p. 14].
#> 
#> ## References

Explanation: Pandoc's citeproc reads citations of the form [@key] and emits the formatted version per your CSL stylesheet; the page suffix after the comma renders as "(Smith 2020, 14)" or similar. The references section auto-populates at the end of the document, so the convention is to put your ## References heading near the bottom and pandoc fills it. To cite multiple works at once: [@smith2020; @jones2021].

ex_5_2 <- "As shown in Figure \\@ref(fig:mpg-wt), heavier cars consume more fuel."
cat(ex_5_2)
#> As shown in Figure \@ref(fig:mpg-wt), heavier cars consume more fuel.

Explanation: Bookdown prepends fig:, tab:, or eq: to the chunk label depending on the artifact type, which is why the cross-reference always specifies the prefix. Cross-references only resolve under the bookdown::html_document2 (or pdf_document2, word_document2) output formats, not plain html_document. If a reference renders literally as \@ref(fig:mpg-wt) in the output, the chunk almost certainly lacks fig.cap: bookdown needs a caption to register the label.

ex_5_3 <- "## Model evaluation {.tabset}\n\n### Accuracy\n\n### Calibration\n\n### Drift"
cat(ex_5_3)
#> ## Model evaluation {.tabset}
#> 
#> ### Accuracy
#> 
#> ### Calibration
#> 
#> ### Drift

Explanation: The .tabset class turns every immediate child heading into a clickable tab, with the first tab active by default. Add .tabset-pills for pill-style buttons, or .tabset-fade to fade between tabs. To close the tabset, drop back to the parent heading level or higher; otherwise every later subheading also becomes a tab, which is the most common surprise when the section runs longer than expected.

ex_5_4 <- '---
title: "Quarterly KPI Review"
format:
  html:
    toc: true
    toc-location: left
    code-fold: true
params:
  region: "EMEA"
---'
cat(ex_5_4)
#> ---
#> title: "Quarterly KPI Review"
#> format:
#>   html:
#>     toc: true
#>     toc-location: left
#>     code-fold: true
#> params:
#>   region: "EMEA"
#> ---

Explanation: Quarto normalises field names to dash-separated lowercase (toc-location, code-fold) versus Rmd's toc_float, code_folding; both engines accept either convention but Quarto code in the wild uses dashes. The top-level key is format: (not output:), and the HTML-specific options live one level deeper under html:. Quarto's params: block accepts plain key-value pairs without the value: wrapper Rmd needs for typed defaults.

regions <- c("EMEA","APAC","NA")
ex_5_5 <- vapply(regions, function(r) {
  rmarkdown::render(
    "report.Rmd",
    output_file    = paste0("report_", r, ".html"),
    params         = list(region = r),
    output_options = list(pandoc_args = c("--metadata", paste0("title=Q1 KPI Review: ", r))),
    envir          = new.env()
  )
}, character(1))
ex_5_5
#> [1] "report_EMEA.html" "report_APAC.html" "report_NA.html"

Explanation: Pandoc reads metadata from both YAML and command-line --metadata key=value pairs, with command-line winning, which is how you override the static YAML title without editing the source. output_options is forwarded to the underlying output-format function (html_document() here), so pandoc_args lands in the right place. For a richer override (subtitle, author, date) chain multiple --metadata pairs in the same vector.