0

I'm confused by the exercise solutions of the book Applied Predictive Modeling. In

https://github.com/topepo/APM_Exercises/blob/master/Ch_06.pdf at the beginning of page 4

Resampling results
  RMSE   Rsquared  RMSE SD  Rsquared SD
  1.382  0.8716    1.014    0.08729
The RMSE is slightly more than 1.4%

To my knowledge RMSE isn't a % (of what?) It's

mse <- mean(residuals(fit)^2)
rmse <- sqrt(mse)

Can someone help me on this?

Hong Ooi
  • 7,629
  • 3
  • 29
  • 52
P6000
  • 1
  • 1
  • 1

1 Answers1

1

The Root Mean Squared Error (RMSE) is calculated on the original scale. If the data to be predicted is in kilograms, your RMSE will be expressed in kilograms. If your data are in degrees Fahrenheit, your RMSE will be in degrees Fahrenheit.

In this specific case, your data are the fat content as a percentage of meat. Your data are percentages. So your RMSE will also be expressed in percent.

Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357