zmjones/mmpf

Monte-Carlo methods for prediction functions

mmpf's Issues

marginalPrediction calls 'predict' changes order of data columns before calling predict

See mlr-org/mlr#2439. makeDesign changes the order of data columns:

> colnames(iris)
[1] "Sepal.Length" "Sepal.Width"  "Petal.Length" "Petal.Width"  "Species"
> colnames(mmpf::makeDesign(iris, "Petal.Width", dim(iris)))
[1] "Sepal.Length" "Sepal.Width"  "Petal.Length" "Species"      "Petal.Width"

which is fine by itself. However, when it is used in marginalPrediction, the predict function is then called with this reordered data, which breaks in some cases where predict expects the order of columns to match the order during model fitting. A fix for this would be a design = design[colnames(vars)data)], or a change in makeDesign that preserves column order.

Resolving Error: Assertion on 'x' failed: Must be of type 'data.frame', not 'logical'.

I'm getting the following error, and it's not clear how to resolve this:

Error in marginalPrediction(sample.features[, -c("revenue_log_s")], "empcount_s",  :
  Assertion on 'x' failed: Must be of type 'data.frame', not 'logical'.

I successfully fit a randomForest model, and then call marginalPrediction like so:

mp = marginalPrediction(data[, -c('revenue_log_s')], 
                        "empcount_s", 
                        c(10, nrow(data)), 
                        fit)

I'm not sure what 'x' is referring to, as the data I pass in the first row is a dataframe, and I'm sure that the second argument is supposed to be a character vector, and not a dataframe.

Any thoughts?