3
$\begingroup$

I am composing a regression tree to determine corporate dividends with dividends as the determinant and 13 predictor variables. To check the accuracy of the tree I am using RMSE values computed using a random sample set. The problem is my RMSE values are huge since the range of dividends (the determinant) is so large.

I was wondering if I should scale the determinant or maybe all of the input data? If I scale the data I get a smaller RMSE but the tree is difficult to read since it has scaled values. Thanks.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

Scaling the variables doesn't change the amount of error, it just changes units. Regression trees are, I am fairly sure, insensitive to monotonic transformations of the predicted variable, so you will get the same result.

Improve this answer
$\endgroup$
1
  • $\begingroup$ You can be almost sure :) $\endgroup$
    – Momo
    Commented Oct 3, 2012 at 20:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.