A bug regarding the predictions generated by rf() that affected every other function fitting models has been fixed. Previously, the model predictions came from the "predictions" slot produced by ranger(). Such predictions are produced from the out-of-bag data during model training, and are different and lead to lower R squared values than those produced with predict(). Now the predictions yielded by rf() are generated with predict(), and therefore you might notice that now models fitted with spatialRF functions perform better than before, because they do.
The function print_evaluation() does not use huxtable any longer to print the evaluation results, and only shows the results of the testing model.
Added support for binary data (0 and 1). The function rf() now tests if the data is binary, and if so, it populates the case.weights argument of ranger with the new function case_weights() to minimize the side effects of unbalanced data.
Fixed an issue where rf() applied the wrong is.numeric check to the response variable and the predictors that caused issues with tibbles.
Removed the function scale_robust() from rf(), and replaced it with scale(). It was giving more troubles than benefits.
Simplified rf_spatial().
Modified rf_tuning() to better tune models fitted with rf_spatial().
Minor fixes in several other functions.
All 'sf' dependencies removed from the package.