Tidy a "train" object using broom

user_13628719 0
Viewed 216

I made a linear regression model by using the caret package with the code below

library(caret)
#Anscombe data is available on R
model_1<-train(
  form=y1~x1,
  data=anscombe,
  method='lm',
  trControl=trainControl(method='cv',number=3))

What I wanted to do is convert the model into a data frame using broom::tidy(model_1), but it throws an error

# Error: No tidy method for objects of class train

I think the problem is with the class of the caret's model, which is train() instead of lm(). Is there any way to tidy a train object? Or should I convert the train object into lm first?

broom r-caret linear-regression machine-learning r
1 Answers

This type of object is not currently supported by the broom package. See here: https://github.com/tidymodels/broom/issues/128

However, you could easily define your own tidy method by following the instructions here: https://www.tidymodels.org/learn/develop/broom/

Here is a minimal example:

library(caret)
library(broom)

tidy.train <- function(x, ...) {
  s <- summary(x, ...)
  out <- data.frame(term=row.names(s$coefficients),
                    estimate=s$coefficients[, "Estimate"],
                    std.error=s$coefficients[, "Std. Error"],
                    statistic=s$coefficients[, "t value"],
                    p.value=s$coefficients[, "Pr(>|t|)"])
  row.names(out) <- NULL
  out
}

model_1<-train(
  form=y1~x1,
  data=anscombe,
  method='lm',
  trControl=trainControl(method='cv',number=3))

tidy(model_1)
#>          term  estimate std.error statistic     p.value
#> 1 (Intercept) 3.0000909 1.1247468  2.667348 0.025734051
#> 2          x1 0.5000909 0.1179055  4.241455 0.002169629
user_3423310
Related