<!---searched in commits on statcheck repo: `committer-date:2020-03-15..2020-04-30`
and checked out changes in the beta releases --->
Major external changes
* The variable names from the output of `statcheck()` have changed to increase consistency in style and naming. This means that the variable names in the output of `checkPDF()`, `checkHTML()`, `checkdir()`, `checkPDFdir()`, and `checkHTMLdir()` have also changed.
Major internal changes
There have been major updates to the internal structure of statcheck. Some of the most notable are:
* The main `statcheck()` function has been significantly shortened by summarizing a lot of repeating actions into general functions, and by moving functions from within `statcheck()` to their own scripts.
* Combine the regular expressions for all separate tests into one main regex and write generic functions to parse the extracted NHST results. In the previous version, a lot of the regexes overlapped, and many actions were performed multiple times throughout the code, which made the script inefficient and error-prone
* Update the way that errors and decision_errors are determined. Now, checking for correct rounding is done within the error function, instead of in a separate function
* Calculate one-tailed p-values more sophisticatedly: instead of simply doing p/2, statcheck now actually calculates a one-tailed p-value based on the surface under one tail of the appropriate distribution. This ensures that for one-tailed p-values, correct rounding is also taken into account
* All documentation and the NAMESPACE are now generated with roxygen2
* The variable names in the output are now based on a file with constants. This makes it easier to update the names in a later stage if necessary, without having to go through every script
Small updates
* Don't show a message to warn for the potential presence of one-tailed tests and other significance levels. This text was mainly distracting.
Bug fixes
* accurately take correct rounding into account with negative test statistics
* extract punctuation that could signal a wrongly encoded minus sign
* added additional html encodings of mathematical symbols
* ignore result when test value == NA
* test results with multiple comparison signs are no longer extracted (e.g.: "t(38) >= 2.25, p = .03")