NEWS.md
tidy_stats() on ungrouped count data produced with count_data() is now properly tidied.tidystats does not support a particular analysis, you can create your own generic test by providing a list of statistics.anova().BayesFactor functions.pkgdown website for the package.tidy_stats_to_data_frame function, and a description of the tidystats taxonomy.anova().count_data() again.read_stats() now converts Inf character strings to numeric.write_stats() now has a digits argument that determines the number of decimals for saved numbers (default: 6).describe_data() caused by the dplyr 1.0.0 update.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().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.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.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).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.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.report_statistic(). Consequently, all report functions have been updated to use this new generic function.identifier column from each list element when using read_stats().tidy_stats.lm() and tidy_stats.glm() to be consistent with the other tidy_stats() functions.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).glm().lmer() and lmerTest’s lmer().alpha().ICC().confint() via the new class argument in add_stats() and add_stats_to_model().add_stats_to_model().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.read_stats() now removes empty columns from each list element.describe() to describe_data() so that it no longer conflicts with psych’s describe().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.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.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.report() you can use M() and SD() to quickly report the mean or standard deviation, without having to specify that particular statistic. Less typing!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!describe_data()
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.chisq.test and wilcox.test.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”).report(). The function will now throw an error when the identifier does not exist.report_p_value() to support multiple p-values.htests were always 95% confidence intervals.aov() output.add_stats_to_model(). The method previously required a term and did not automatically complete information (e.g., method information).