`tidystats`

has been completely redesigned in terms of how statistics are combined together. While previously the output of statistical models was converted to a tidy data frame, the output is now converted to a list, with an entirely different structure. The reason for this change is that lists are more machine-readable, enabling more interesting features down the line. It is still possible to convert the list of statistics to a single data frame with a new function called`tidy_stats_to_data_frame()`

.- The significant changes made to
`tidystats`

has resulted in the loss of some previously supported statistical functions. For a list of currently supported statistical functions, see the help document of`add_stats()`

or the README. - All
`report`

functions have been removed for now. These may return (if I get the impression these are liked) but for now I am focusing my development time on creating a Word add-in that will enable researchers to use a`tidystats`

-produced file for reporting statistics in Microsoft Word. `describe_data()`

no longer accepts multiple column names. The goal of the function is now to calculate the descriptives of a single column (which can still be grouped to calculate the descriptives for each group level).`count_data()`

has been removed.

`add_stats()`

now has a`type`

argument to specify whether an analysis was a primary analysis, secondary analysis, or exploratory analysis.`add_stats()`

now has a`preregistered`

argument to specify whether an analysis was preregistered or not.

- Added an example dataset called ‘quote_source’ containing data of a replication of Lorge & Curtiss (1936) that was part of the Many Labs project (Klein et al., 2014)

- Changed the argument order in the family of
`add_stats()`

functions. Previously, the model output or tidy data frame was the first argument. This allowed you to directly pipe the model output into`add_stats()`

(using**magrittr**’s %>%). However, an alternative approach is to have the tidystats list to be the first argument. This allows you create a long sequence of pipes. You start with the results list, add a model via`add_stats()`

, pipe the result into the next`add_stats()`

, and so on. Since you often store your model output in variable names anyway, this is probably more convenient. Additionally, this probably also keeps your script more tidy (you can do this at the end of your data analysis script). - Certain statistical models are now tidied differently due to the addition of a ‘group’ column. Several models like multilevel models, meta-analytic models, and arguably also regression models have more than just terms (e.g., model fit), so to distinguish between coefficients and other parts of the output, a ‘group’ column has been added. This also means usage of the
`report()`

is affected, as now the group should be specified when necessary. Affected models are regression, within-subjects ANOVA, multilevel models, and meta-analysis models. - Added the
*class*argument to`add_stats()`

and`add_stats_to_model()`

. Rather than having to manually tidy the data first, you can make use of some custom`tidy_stats()`

functions by specifying the class argument. Run`?add_stats`

to see a list of supported classes and see the help document of`tidy_stats.confint()`

for an example. - Under the hood: Added a generic report function for single values called
`report_statistic()`

. Consequently, all report functions have been updated to use this new generic function. - Removed the
`identifier`

column from each list element when using`read_stats()`

. - Reordered the columns of
`tidy_stats.lm()`

and`tidy_stats.glm()`

to be consistent with the other`tidy_stats()`

functions.

- Added a new function called
`inspect()`

. This function accepts a tidystats results list or the output of a statistical model and will display all results in RStudio’s Viewer pane. This allows the user to visually inspect the results and, importantly, copy results in APA style to their clipboard. This function is aimed at users who prefer not to use R Markdown or when you want to quickly run a model and get the results in APA-style. This new function works well with Microsoft Word, but does not work with Apple Pages (some of the styling is lost when copying the results). - Added support for
`glm()`

. - Added support for lme4’s
`lmer()`

and lmerTest’s`lmer()`

. - Added support for psych’s
`alpha()`

. - Added support for psych’s
`ICC()`

. - Added support for stats’
`confint()`

via the new`class`

argument in`add_stats()`

and`add_stats_to_model()`

.

- Added check for an existing identifier in
`add_stats_to_model()`

. - Added a
`class`

argument to`add_stats()`

and`add_stats_to_model()`

. Some statistical tests return a normal data.frame or matrix, which does not specify which test produced the results. This makes it difficult for tidystats to figure out how to tidy the result. Previously, we solved this by`add_stats()`

accepting pre-tidied data frames. Now we added a the`class`

argument to specify the name of the function that produced the results, so that we can then tidy it for you. - Added warnings in case unsupported output is added (e.g., a pre-tided data frame).
`read_stats()`

now removes empty columns from each list element.- Improved documentation.

- Fixed a bug that would incorrectly classify ANOVAs as One-way ANOVAs when character variables were used rather than factors.
- Prepared for
`dplyr`

0.8

- Added tests to the R package to minimize bugs.
- Made the code and documentation more consistent
- Added an under-the-hood helper function to rename statistics columns

- Renamed
`describe()`

to`describe_data()`

so that it no longer conflicts with**psych**’s`describe()`

. - Changed
`describe_data()`

to no longer accept non-numeric variables, but added the feature that descriptives can be calculated for more than 1 variable at a time. It is recommended to use the`count_data()`

function for non-numeric variables. - Renamed
`tidy_descriptives()`

to`tidy_describe_data()`

and improved the function. A notable change is that var information is now returned to identify which descriptives belong to which variable. Also changed the group delimiter to ’ - ’. `write_stats()`

now prettifies the numbers using`prettyNum()`

when saving them to disk.

- Improved
`report()`

function. The method now supports the option to retrieve a single statistic from any tidy stats data frame. This will allow you to report all statistics, even when reporting functions for a specific method are not yet supported. - Added quick report functions for means and standard deviations. Instead of using
`report()`

you can use`M()`

and`SD()`

to quickly report the mean or standard deviation, without having to specify that particular statistic. Less typing! - Added an option called ‘tidystats_list’ in
`options()`

to set a default list. By setting the tidystats list in`options()`

, you do not need to specify the list in the**results**argument of`report()`

. Less typing! - Reporting regression results will now include a check for whether confidence intervals are included, and report them.
- Added skewness and kurtosis to
`describe_data()`

- Added new
`count_data()`

function to calculate count descriptives of categorical data. Also added a`tidy_count_data()`

function to tidy the output of this new function. - Added support for
`chisq.test`

and`wilcox.test`

. - Added a better default
`identifier`

to`add_stats()`

. If you supply a variable to be added to the tidystats list, and no identifier is provided, it will take the variable name as the identifier. If you pipe the results into`add_stats()`

then the old default identifier will be used (e.g., “M1”).

- Added identifier check to
`report()`

. The function will now throw an error when the identifier does not exist. - Added statistic check to all report functions. The function will now throw an error when the statistic does not exist.
- Improved
`report_p_value()`

to support multiple p-values. - Updated documentation to be more consistent and to take into account the changes made in the current update.

- Fixed bug that it was assumed that confidence intervals in
`htests`

were always 95% confidence intervals. - Fixed bug in report functions that would occur when no statistic argument was provided.
- Removed spaces from terms in
`aov()`

output. - Removed a leading space from the method information of a Two Sample t-test.
- Improved
`add_stats_to_model()`

. The method previously required a term and did not automatically complete information (e.g., method information).

- Initial release