NEWS.md
A new function report_model was introduced to allow simple model reports. See ?report_model for details and documentation.
We now use testthat edition 3 which makes it easier to deal with warnings and errors in a cleaner way. See https://github.com/r-lib/testthat/issues/1471, https://github.com/Nelson-Gon/manymodelr/pull/22.
Updated all examples to use the new dataset, yields.
Topic based vignettes are now available.
Added a new dataset yields that may be useful for testing purposes.
Fixed issues with knitr causing failing builds.
Updated docs with newer examples.
fit_models to support model fitting for several variables for several model types.Major additions
extract_model_info now supports glmerMod and glmmTMB
get_this now works with numeric input and also supports data.frame objects.
fit_models extends fit_model by building many models at once.
Other changes
get_stats now drops columns via a vector and not “non_numeric” as previously.
Metrics from multi_model_1 are now more informative with the metric and method wrapped in the naming of the result.
df was renamed as old_data in multi_model_1, newdata to new_data.
plot_corr now directly accepts data.frame objects. Arguments like round_values have also been dropped.
Fixed DOI to Max Kuhn’s paper
Refactored get_mode to be tidy compliant.
The argument valid was dropped in multi_model_1.
get_all was dropped in select_percentile.
select_col, select_percentile, row_mean_na will be removed in the next release.
row_mean_na is now defunct. Use na_replace instead.
na_replace no longer allows using functions such as mean,min, etc. These have been reimplemented in the package mde
modeleR is now defunct. Use fit_model instead.
get_this no longer accepts non quoted character strings.
Better coverage and code tests
New functions
plot_corr has been added to allow plotting of correlation matrices produced by get_var_corr_.
na_replace_grouped extends na_replace by allowing replacement of missing values(NAs) by group.
add_model_predictions allows addition of predicted values to a data set.
add_model_residuals is an easy to use and dplyr compatible wrapper that allows addition of residuals to a data set.
extract_model_info allows easy extraction of common model attributes such as p values, residuals, coefficients, etc as per the specific model type. It supports extraction of multiple attributes.
multi_model_2 allows fitting and predicting in one function. It is similar to multi_model_1 except it does not require metrics.
Major Changes
modeleR has been replaced with fit_model which is an easier to remember name. Usage remains the same.
fit_model no longer allows direct addition of predictions. Use add_model_predictions to achieve the same.
na_replace has been extended to allow for user defined values.
rowdiff now accepts replacement of the calculation induced NAs. It does so by using na_replace.
get_var_corr_ now supports using only a subset of the data.
Helper functions are no longer exported.
get_data_Stats is now aliased with get_stats for ease.
get_var_corr no longer has the get_all argument. Instead, users can provide an option other_vars vector of subset columns. drop_columns has also been changed from boolean to a character vector.
Major Changes
Additions
agg_by_group is a new function that manipulates grouped data. It is fast and robust for many kinds of functions.
rowdiff is another new function that enable one to find differences between rows in a data.frame object. `
get_var_corr provides a user-friendly way to find correlations between data.
get_var_corr_ provides a user-friendly way to find combination-wise correlations. It is relatively fast depending on how big one’s data is and/or machine specifications.
get_this is an easy to use helper function to get metrics,predictions, etc. Currently supports lists and data.frame objects.
modeleR and row_mean_na were removed.
Major Modifications
get_data_Stats now supports removal of missing data as well as using only numeric data.
modeleR has been fixed to handle new data as expected. It also now supports glm.
multi_model_1 now supports either validation or working with new data.
row_mean_na has been replaced with na_replace which is more robust. row_mean_na will be removed in future versions.