NEWS.md
correlation() gains a missing= argument, similar to stats::cor(use=), for controlling how missing data is handled.
correlation() converts numeric input variables automatically into factors when method = "polychoric".
The format() method for objects of class easycormatrix gets a zap_small argument, to round very small numbers.
cor_sort() can now deal with non-square matrices.
Updated required R version to >= 4.1.0 (released May 2021) to pass CRAN checks on documentation
cormatrix_to_excel() function for exporting correlation matrices to Excel with color formatting.Minor improvements and code revisions due to changes in other packages.
Default color scheme for correlation matrices switched to use red for negative values and blue for positive values.
distance_mahalanobis() is deprecated. Use performance::check_outliers(method = "mahalanobis_robust") instead.
The minimum needed R version has been bumped to 3.6.
Fixes breakages caused by updates to parameters package (#269).
The visualization recipe (plots) for redundant correlation matrices was improved, so self-correlations will no longer be labelled and get a neutral color.
The print() method redundant correlation matrices no longer shows the diagonal with self-correlations.
plot() methodswinsorize() function, which now lives in datawizard package.cor_smooth() function for smoothing non-positive definite matrices.data2 was specified correlation() was over-correcting for all of the combinations of variables in the full x and y tables, rather than in just the ones specified (#195).correlation() gains a new argument rename to rename variables.
simualte_simpson() function is now re-exported from bayestestR package.
plot() for "easycor_test" objects now produces an annotated scatter plot.
simualte_simpson(): The groups are now named after the pattern "G_" (can be altered with the group_prefix argument).
robust argument deprecated in favour of ranktransform.
correlation gains two new arguments: select and select2 to select specific variables from dataframes to compare (#146).
as.matrix method works for grouped correlations (#148).
New as.list method returns a list of various matrices related to correlation analysis (correlation, number of observations, p-values, etc.).
0.6.0 release introduced a bug in Winsorized Pearson correlation where the missing values were removed from the entire data, instead for each pair (#151). This is now fixed.Added verbose arguments to some functions, to toggle warnings on/off.
cor_test() (and hence, correlation()) now default the winsorize argument to .1 when it’s set to TRUE.
The Method column in output dataframe is now more explicit about the correlation method used.
NAs are present (#130).Added winsorize() function.
Added winsorize argument for Winsorized correlations.
Added method = "somers" to correlation(), to compute Somers’s Dxy rank correlation for binary outcomes.
New function display(), to print output into different formats. Currently, only markdown is supported. print_md() is an alias for display(format = "markdown").
cor_to_p() that gave slightly different test statistics.Don’t error if less than 3 valid observations (#100).
Add “gaussian” rank method.
Add “gamma” method.
Add “hoeffding” method.
Add “blomqvist” method.
Added Method column to Bayesian correlations.
Fix bug when robust=TRUE (#87).
Removed CI threshold column from distance_mahalanobis()
Fixed bug (#76)
Fixed issue in correlation() for some edge cases when include_factors = TRUE.
Fixed issue in correlation() for correlation coefficients with less than four complete pairs of observations (in such cases, cor_test() now returns NA for the confidence intervals).