Passing parameters to dots that conflict with explicit parameters

Question

The caret::train() function has an explicit method parameter, for which we can specify the machine learning method to use (like MASS::polr). Further, caret::train() allows you to pass parameters to the method function via the ... (dots) parameter.

However, one of the MASS::polr parameters I'd like to pass is method, which conflicts with the caret::train() method parameter.

How does one specifically pass the polr method="probit" parameter to polr while still passing the method="polr" parameter to caret::train()?

Not sure if there is a specific trick for caret::train() or a general trick for function dots vs. named function parameters that clash in general.

Specifically, I need:

  ## Generic example, no data, but captures idea
fit <- train(xdata, ydata, method="polr" # this 'method' is named parameter for train() function
  , preProcess = c("center", "scale")
  , method="probit" # this 'method' parameter needs to be passed via dots to 'polr'
  )

Answer 1

In general the ... option passes the needed parameters to the underlying function. But in some cases this indeed clashes with the existing parameters in the train function. In case of polr this is solved via tuneGrid.

See the available models page and search for polr.

In the formula notation it should look like this:

train(y ~ x1 + x2, 
      data = my_data, 
      method = "polr", 
      trControl = my_control, 
      tuneGrid = expand.grid(method = "probit"))