Changelog

Updated code_allophones
- MOUTH gets a prelateral allophone now: PROWL. This means the elsewhere allophone is now called BOUT.
- PRICE gets a prelateral and a prerhotic allophone: CHILD and PRIOR.
- Prevoiceless PRICE has been changed from PRICE to BITE. Prevoiced PRICE has been changed from PRIZE to BIDE.

Fixed two issues with GOOSE allophones in code_allophones.
- Words like tool which are both post-coronal and prelateral, were previously coded as post-coronal. That has been fixed so that they’re not prelateral.
- There wasn’t a prerhotic allophone of GOOSE, which caused words like demure to be erroneously classified as TOOT. Now, prerhotic GOOSE is CURE. Note that prerhotic FOOT is also CURE.

norm_logmeans() has a raised cap before triggering a warning message.
find_outliers() now as the verbose argument which, by default, suppresses the warning message triggered when there is not enough data.

code_allophones() now added.
norm_logmeans() now added, which does a log-means normalization based on Barreda & Nearey (2018).
norm_anae() updated. The use_telser_g argument has been changed to simply g. Previously, if you wnated to calculate g internally, you’d do use_telsur_g = FALSE and that code didn’t work. Now, you’d say g = "calculate", and that does work this time. Additionally, some new functionality is that g can be set to any arbitrary value, such as zero: g = 0.

arpa_to_keywords() has been replaced with the more robust and flexible switch_transcriptions.

norm_anae() now uses the Telsur G by default. This can be suppressed and a new G value can be calculated using the data (the old default behavior), by specifying use_telsur_g = FALSE.
norm_anae() now requires the data to be grouped by speaker prior to running. This makes it more consistent with the other normalization procedures in this package.
wells_to_arpa() was created, as an inverse function to arpa_to_wells() because I needed it one time.
norm_deltaF() has been fixed so that including F4 now works. Additional arguments have been added as well.

pillai() and tidy_mahalanobis() get much clearer examples and much more documentation.
arpa_to_keywords() no longer crashes if a dataset didn’t have all vowels present.

Switching to pkgdown, which involved some overdue tidying of the package
- Removed depreciated functions, including hz, bark, spread_n, move_x_before_y, move_x_after_y, and all the functions related to the old outlier detection procedure (joey_filter, joey_do_pca, joey_adjust_cutoff, joey_rm_outliers).
- Removed the datasets. They were just added like a week ago, but I decided to move them to a separate package joeysvowels.
- Renamed the .R filenames
Added a NEWS.md file to track changes to the package.

joeyr 0.9.2 (May 1, 2025)