We can include levels with no observations in dplyr::count() now with the .drop = FALSE argument setting. I changed the wording and added an example. Please feel free to amend text wording and remove example if you wish.
In transform.Rmd (Chapter 5, Data Transformation), under Counts (5.6.3),
we removed na.rm = TRUE when computing mean because all NAs for
arr_delay were already removed when creating not_cancelled dataset.
* No need to load ggplot2, tidyverse is loaded
* Add motivation for why to pivot + minor edits
* Add an exercise on making a contingency table
* Use index.Rmd instead of rmd to make the build work
* Case study should be sentence case
* Make figure more legible and don't rely on color alone
* Streamline example + names_transform
* Potential pathway for more advanced pivot_wider
* Pivot even wider example + new unite exercise
* Update case study to use full feature pivot_wider
* Change the pregnancy exercise
* Fix chapter reference
* Remove century example since first 2 digits is not century
* Add caption to floating figure
* Align parens
* Fix referencing
* Add alt text to all chunks with plots or figures
* Remove unused figure (related to year -> century)
* Make subparts of exercises alphabetical
* Remove references to names_transform, use mutate()
* Fix code style
* Few more chp/sec reference fixes
* Put line break back
The `..` syntax in 3.7 Exercises to refer to computed variable `prop` should be updated to use the best practice of `after_stat` instead. Reduces confusion and more aligns with the documentation.
Line 506:
When fixing the string format in the `key` column using the `str_replace` function, I think the column name was mistakenly stated as "names_from", where it should actually be "key", which is the original column name.
Further to {tidyselect} documentation "one_of() is retired in favour of the more precise any_of() and all_of() selectors." this exercise (5.4.1) should be updated to be in-line with current functionality.
