Error in case_res[i, , drop = FALSE] : incorrect number of dimensions

[Mateusz Staniak]

I'm trying to use lime with a 1-d model as a toy example and I'm getting this error.
(Based on an example in Christoph Molnar's IML book)

set.seed(17)
x <- runif(1000, -8, 8)
y <- ifelse(x <= 1, (x + 3)^2 - 10 , -5*x + 11)
molnar <- data.frame(x = x, y = y)

true_model <- function(model, newdata) {
  ifelse(newdata$x <= 1, (newdata$x + 3)^2 - 10 , -5*newdata$x + 11)
}

true_model(list(), data.frame(x = c(0, 10)))

library(lime)
lime_wr <- lime(molnar, list())
model_type.list <- function(x, ...) "regression"
predict_model.list <- function(x, newdata, type, ...) true_model(x, newdata)

lime_expl <- lime::explain(data.frame(x = -6, y = 2), lime_wr, n_features = 2)
# Error in case_res[i, , drop = FALSE] : incorrect number of dimensions

[Thomas Lin Pedersen]

lime expects the output to be a data.frame, not a vector. see ?model_support for how to format the output of true_model()