Error using RFE and Recipes package

[mmarcato]

I'm trying to implement PCA and run RFE to select the best components. I used the recipes package for pre-processing and added that model recipe to the rfe. However, I get the following error "Error in eval(e, x, parent.frame()) : object 'Num_Resamples' not found". See the code below:

df <- twoClassSim(100)

# Create recipe for pre-processing
model_recipe <- recipe(Class ~ ., data = df )  %>%
        step_impute_mean(all_numeric()) %>%
        step_normalize(all_numeric()) %>%
        step_pca(all_predictors(), threshold = .95)

# Set ROC as the metric for the Logistic Regression function
lrFuncs$summary <- twoClassSummary
set.seed(42)

ctrl <- rfeControl(functions = lrFuncs,     # Logistic Regression
                    method = "LOOCV",             # Leave One Out Cross Validation
                    verbose = FALSE)

# Recursive Feature Elimination
rfe_lr <- rfe(model_recipe , 
                data = df, 
                sizes = c(2:12),
                rfeControl = ctrl,
                metric = "ROC",
                maximize = TRUE)

I think the problem is something related to using LOOCV as if I try using some other validation methods, it does run. I think that I need to specify Num_Resamples. After looking through the documentation and searching for this error, I tried different things, but nothing worked. I have two questions:

1. Is there any argument missing in my implementation with the recipes package? My understanding is that the argument number does not need to be specified when using LOOCV.
2. Alternatively, would it be possible to implement this using the trainControl (preProcOptions argument)? My understanding is that trainControl only works with the train function and not with rfe function - I've seen some example codes passing trainControl to the rfe function, but my tests indicate that rfe ignores the argument trControl.