NEWS.md

Crosstables for descriptive analyses. See documentation at https://danchaltiel.github.io/crosstable/ and browse code at https://github.com/DanChaltiel/crosstable.

crosstable 0.8.1

Bug fixes and improvements

• fixed as_flextable(compact=TRUE) with duplicated labels (#87).

crosstable 0.8.0

New features

• Remove percents when N=0 with crosstable(remove_zero_percent=TRUE) (#78).
• Better header for multi-by flextables as_flextable(ct, remove_header_keys=TRUE) (#74).
• Control font size with body_add_normal(font_size) and crosstable_options(normal_font_size) (#65).
• Option control on height & width in body_add_gg2() (#68).

Bug fixes and improvements

• Add envir argument to functions that call glue() (#83).
• Better error when x is NULL in body_add_table_section() (#82).
• Warn if by is "" (#77).
• Default to quiet warnings if columns are missing (#76).
• Moved the legend above the table in body_add_table_section() (#73).
• Fixed get_label() on named vector columns (#72).
• Fixed crosstable(drop_levels) (#69, #70, #71).
• Flextable autofit now uses the recommended method (#62).
• transpose() will work on multi-by crosstables (#60).
• Fixed a bug happening when NA is already a level (#59).
• Fixed error where by contains "label" (#56).
• Fixed as_flextable(header_show_n) not working in single by (#54).
• Fixed some missing options in crosstable_options() and added tests (#53).

crosstable 0.7.0 _^2023-11-12

New features

• New officer helper body_add_table_section() that adds a table (crosstable or flextable) which can be surrounded by a title (level 3 by default), a legend, and/or a short sentence.
• set_label() now accepts a function as a value (e.g. set_label(mtcars2, toupper)).
• Argument format_fixed(scientific) can now be FALSE to force standard format. Can be set through options(crosstable_scientific_log=FALSE) (#49).
• New argument format_fixed(epsilon) to format values as "<epsilon". Can be set through options(crosstable_format_epsilon=0.001).
• Markdown implementation has improved, with support of crossing patterns (#26), newline tokens (#30), and many new tags (#51).
  You can now use body_add_normal("Here is **`some code` in bold & *italic* ** <br> And here is <color:red>red text</color>.")
  See ?body_add_normal for more insight.

Bug fixes and improvements

• Reimplement description of columns containing both NA (missing values) and "NA" (characters) (#42).
• Argument as_flextable(by_header) can now be set through options(crosstable_by_header=FALSE) to remove all headers.
• Argument body_add_crosstable(header_fontsize) now defaults to 1.2*body_fontsize.
• Adds an error message in as_flextable() when the crosstable is empty (#41).
• Fixes a regression where references were not added anymore.

crosstable 0.6.2 _^2023-05-26

Bug fixes and improvements

• Fixes for CRAN checks.
• Fixes write_and_open() not opening sometimes.
• Trailing commas will not make crosstable_options() fail anymore.

crosstable 0.6.1 _^2023-04-03

Fixes for CRAN checks.

crosstable 0.6.0 _^2023-03-26

New features

• New function transpose_crosstable() (or simply t()), which transposes a crosstable so that by is in rows instead of columns.\

  ct = crosstable(mtcars2, c(mpg, drat), by=am)
  t_ct = t(ct)
  as_flextable(t_ct, by_header="Variables")

• New function pivot_crosstable(), which pivots a crosstable so that variable is spread as several columns.\

  ct = crosstable(mtcars2, c(mpg, drat))
  p_ct = pivot_crosstable(ct)
  as_flextable(p_ct)

• body_add_table_list() now replaces body_add_crosstable_list() and body_add_flextable_list(). It allows both crosstables and flextable (and even dataframes, which are turned to flextables first) and has a new argument fun_after to control what comes after the table. See examples in ?body_add_table_list.

• New function get_percent_pattern(). See also the new vignette vignette("percent_pattern").

• New argument crosstable(drop_levels=TRUE) to drop unused levels in factors.

• copy_label_from() now works on dataframes as well.

• as_flextable(header_show_n_pattern) can now be a list of names cell and total, so that the "Total" column can be labelled too.

Breaking changes (minor)

• showNA="no" is now consistent with stats::addmargins(), gtsummary::tbl_cross(missing="no"), and janitor::tabyl(show_na=FALSE). It now actually removes all NA from the equation, instead of not doing much (#24).

• In percent_patern, the proportion relative to the total sample p_cell has been renamed to p_tot for clarity.

Bug fixes and improvements

• Removed unexpected warning "NaNs produced" thrown when calculating percentages in totals while the number of NA is higher than the other classes (#20).
• The output is now correct when header_show_n=TRUE and remove_header_keys=TRUE while using multiple by variables (#21).
• Flextables will not have missing columns when using multiple by variables that has same key levels.
• String interpolation works as intended in body_add_xxx_legend().
• crosstable() will not fail if fisher.test() fails [#28]
• forcats::fct_explicit_na() is not used anymore [#29]
• body_add_normal() now removes ` symbols when showing code [#31]
• Trailing commas will not make apply_labels() fail anymore [#32]
• Performance improvement (around 30% for small tables) as confidence intervals are not calculated anymore when not needed [#34]

Many thanks to Stephan Daus (@sda030) for his bug reports and feedback on this release.

crosstable 0.5.0 _^2022-08-16

New features

• New clean_names_with_labels() which cleans the names of the dataframe but keeps the old names as labels. Obviously inspired by {janitor}.
• New variables n_col, n_row, and n_tot available for percent_pattern. Also, every variable has now its counterpart with the _na suffix which accounts for missing values.
  For instance, one can now write:

    crosstable(mtcars2, cyl, percent_pattern="{p_col} ({n}/{n_col}) [95%CI: {p_col_inf}; {p_col_sup}]")
    crosstable(mtcars2, cyl, percent_pattern="{p_col_na} ({n}/{n_col_na}) [95%CI: {p_col_inf}; {p_col_sup}]")

• percent_pattern can now be a list of characters with names body, total_row, total_col, and total_all to also control the pattern in other parts of the crosstable than the body.

Improvements

• crosstable_options() does not need the "crosstable_" prefix anymore, so the autocomplete is less clumsy. Options with the prefix will still work though. crosstable_options(reset=TRUE) is deprecated in favor of the new crosstable_reset_options().
• crosstable_test_args() and crosstable_effect_args() now have arguments to easily control the non-default parameters.
• Allow scientific notation for big numbers. Default to numbers for which abs(log10(x))>4. This can be controlled using options, e.g. crosstable_options(scientific_log=5).
• In MS Word, crosstables will now break across pages by default. You can revert this by using body_add_crosstable(allow_break=FALSE) or using crosstable_options(). This is the pendant of keepnext in officer/flextable.
• New argument body_add_crosstable(max_cols=25), which limits the size of crosstables in Word documents. This prevents very large tables to be wrongly included.
• peek() is now usable on non-crosstable objects as well. as_flextable() method will be applied on the object if available, otherwise flextable() will be applied.
• Better error messages in import_labels() when data_label doesn't have the right columns.

Deprecations

• rename_dataframe_with_labels() is now deprecated in favor of the better named rename_with_labels().
• compact() is now deprecated in favor of ct_compact() to avoid the conflict with purrr::compact(). It will be removed in the next version.

Bug fixes

• body_add_crosstable(..., parse=NULL) will now work as intended .
• by_header can replace the "value" column if by has only one stratum, in both as_gt() and as_flextable() (#9).
• percent_pattern can now contain functions that start by "p" (#10).
• percent_pattern is actually taken into account when by has one single level (#13).
• import_labels() now works even if there is a missing value in the dataframe.
• Minor other bugfixes

crosstable 0.4.1 _{^(2022-02-19)}

• Maintenance release to fix CRAN errors.

crosstable 0.4.0 _{^(2022-02-14)}

New features

• You can now use basic markdown formats in body_add_normal(): **bold**, *italic*, _underlined_, and `code`.
• New keep_id argument in compact.crosstable(), which enables as_flextable(compact=TRUE, keep_id=TRUE)
• You can use as_flextable(header_show_n_pattern) to change the glue pattern of these group sizes. The default is {.col} (N={.n}); you can use {.col_key} and {.col_val} when by has multiple strata.
• Functions body_add_table_legend() and body_add_figure_legend() earn an argument legend_prefix. Useful when set through global options, for instance to make every table/figure start with the name of the study.
• crosstable_options() can handle new options for all those new features.

Improvements

• You can now set normal style directly in body_add_normal() (it was only doable through options before).
• Removed the "variable" header in compacted crosstables.
• Dataset mtcars2 is now a tibble, with its rownames as a column named "model".
• as_flextable(by_header=FALSE) now removes the group header (if by has only one stratum).
• as_flextable(header_show_n) now also works if by has multiple strata.
• as_flextable(header_show_n=TRUE) adds group size for all groups.

Bug fixes

• Fixed a bug when numeric variables are treated as categorical (failed if one had a NA value).
• Fixed a bug that caused body_add_normal() to add an extra empty paragraph if there was a reference in the text.
• Fixed a bug so that body_add_normal() can be used without argument.
• Fixed a bug in effect=TRUE when some groups were present in table() but not in glm() due to missing values.

crosstable 0.3.2 _{^(2021-11-27)}

New features

• Global options management is now easier thanks to crosstable_options() and autocompletion. Almost every argument can now be set using options. See ?crosstable_options for further details.
• You can also use crosstable_peek_options() to see which crosstable option is currently set.
• New argument num_digits in crosstable(). It was about time!
• New argument header_show_n for as_flextable(), which adds the group size (N=xx) to the header of the flextable.
• New arguments (and global options) par_before and par_after in respectively body_add_table_legend() and body_add_figure_legend(), which add an empty paragraph before/after the legend (for readability).
• New function for {officer}: body_replace_text_at_bkms(), to replace several bookmarks at once.
• New global option crosstable_options(crosstable_zero_percent=FALSE), which removes percentages whenever n==0 (as it would always be 0%). I should add it as an argument of crosstable() one day...

Improvements

• Better sorting when numeric variables are treated as categorical (before, 10 was sorted before 2).
• Removed the labelled class which caused too many problems for little to no improvement. See larmarange/labelled#111.

Bug fixes

• fixed a bug in crosstable() occurring when one of funs does not have ellipsis (...) and funs_arg contains an unused argument.

crosstable 0.3.1 _{^(2021-11-14)}

New features

• New parameter percent_pattern in replacement of crosstable(margin=x) for better control over proportion format. Introduces the possibility of displaying confidence intervals (using Wilson score method) along with proportions.
• New function body_add_crosstable_list() to add a list of crosstables all at once, separated by some customizable titles. Also works with flextables and plain old dataframes.
• New argument crosstable_padding_v for as_flextable() to manage vertical padding. Also available as the global option crosstable_padding_v.
• New global options crosstable_{arg} for almost all arguments. See ?crosstable_options for more details.

Bug fixes

• fixed a bug in as_flextable() occurring when showNA=TRUE (header row was disappearing)
• fixed a bug occurring very randomly, when rounding caused format_fixed() to return a numeric value ("Error: Can't combine ..1$value and ..2$value .")
• crosstable now shows all unused levels in by when it is a factor
• removed extra bold columns in compacted crosstables displayed as flextables

Internal

• renamed branch master to main
• use a lot more snapshots in tests

crosstable 0.2.2 _{^(2021-10-18)}

New features

• Added support for multiple by! You can now write crosstable(mtcars, c(mpg, gear), by=c(am, vs)) %>% as_flextable().
• Added a macro that can autofit every table in the document at once. This macro can be generated using the function generate_autofit_macro() which creates a file that should then be imported into MS Word.
• body_add_crosstable() gains a padding_v argument to control the vertical padding of all rows.
• body_add_title() and body_add_xxx_legend) gain a glue functionality. You can now write body_add_title("The iris dataset (nrow={nrow(iris)})", 1).
• as_workbook() can now take a named list of crosstables, that will be considered as sheets.
• New parameter percent in format_fixed(percent=TRUE/FALSE) to easily format percentages.

Minor changes

• style is deprecated in body_add_table_legend() and body_add_image_legend() in favor of name_format.
• Changed the behaviour of some effect calculations that were done by column instead of rows. That might change some outputs but not their meaning.
• body_add_normal() now removes duplicated spaces (squish) in its input by default. Use squish=FALSE to override.
• docx_bookmarks2() gains a target parameter.

Bug fixes

• effect calculation now takes into account the reference level (first level of a factor).
• body_add_crosstable() rightly takes body_fontsize and header_fontsize into account.
• Added few more warnings, so that you know what went wrong.

Internal

• burgled 2 functions using burglr::burgle() to avoid dependency: nortest::ad.test() and DescTools::CochranArmitageTest().
• fixes the bug from the breaking change in testthat (#3).

crosstable 0.2.1 _{^(2021-02-07)}

• First version on CRAN
• Improved functions naming in funs, especially with multiple combinations of named and unnamed functions, including lambda or anonymous
• Use simplify=FALSE in get_label() to get a list instead of a vector

crosstable 0.2.0 _{^(2021-02-02)}

Misc

• added lots of global options for easier implementation. See ?crosstable_options for the comprehensive list.
• added label helpers: apply_labels() (inspired by expss's), copy_label_from() and rename_dataframe_with_labels()
• added as_workbook() to export a crosstable as a formatted openxlsx Excel workbook, for copypasting purpose.
• added peek() to open a crosstable in a temporary Word document, as copy-pasting in RStudio's viewer is very limited.
• fixed the bug when a columns contained both "NA" (string) and <NA> (missing).
• fixed the bug where function in funs was not found if declared in another environment.
• numerous other minor bugfixes and internal improvements.

Officer

• added cross-reference functionality to body_add_figure_legend() and body_add_table_legend().
  Use the bookmark argument to set a reference, then write "\\@ref(my_bkm)" inside body_add_normal() to call it.
• added docx helpers to add lists: body_add_list() and body_add_list_item(). These will unfortunately not work with the default officer template.
• added some alternatives for some officer functions:
• docx_bookmarks2(), which list bookmarks found in the header and footer as well
• body_add_img2(), and body_add_gg2(), which win a units=c("in", "cm", "mm") argument
• write_and_open(), an alternative to print() for documents, which tries to open it right away.

Deprecations

• Ellipsis (...) use in crosstable() has been deprecated for a more "tidy" syntax. Write crosstable(mtcars2, c(disp, vs)) instead of crosstable(mtcars2, disp, vs). Ellipsis will be defunct in future v1.0.
• crosstable(.vars=) has been renamed to crosstable(cols=).
• moystd() has been renamed to meansd().
• body_add_glued() has been superseded by body_add_normal(), which inherits all functionalities and more.

crosstable 0.1.5 _{^(2020-08-02)}

• added minimal support for gt tables (with as_gt()) for those who like them better than flextables
• improved working with officer: added body_add_figure_legend() and fontsize options for body_add_crosstable()

crosstable 0.1.4 _{^(2020-07-16)}

• added save_labels() to ease working with dplyr
• added meanCI() an additional summary function to use in crosstable()'s funs argument
• improved support for Date variables
• multiple, numerous bug fixes
• renamed moystd() to meansd()

crosstable 0.1.3 _{^(2020-06-29)}

• Added support for description of Date variables. Format can be specified in funs_arg with the date_format key.
• Removed some dependencies to ease installation

crosstable 0.1.2 _{^(2020-06-10)}

• Effect refactoring: better error/warning handling
• Name sanitation: replacing "." by "_" in function names
• Better error messages
• Bug fixes

crosstable 0.1.1 _{^(2020-06-07)}

New features and behaviors

• Added format_fixed(), rounding with the right number of decimals (including zeros)
• Added import_labels(), which apply labels taken from a source dataframe (name, label) to another dataframe
• Added margin="none" option, to remove percentages and keep only counts
• Columns of unsupported class are dropped with a warning instead of failing with an error

Misc

• Method cross_to_flextable() (ctf()) was deprecated and renamed as_flextable() (#207)
• Reexporting pipes and tidyselect helpers so that user does not have to load these libraries
• Computing time optimization (speed x2.6!)
• Fixed bug in normality testing
• Fixed bug in compact()

crosstable 0.1.0 (2020-04-09)

• First release, big changes from the biostat2 package.