Linear Structure

Classification and regression trees are not particularly useful when it comes to deciphering linear relationships, having no choice but to produce a long line of splits on the same variable. If the modeler suspects strong linear relationships, small trees can first be fit to the data to partition it into a few more similar groups, and then standard parametric models can be run on these groups. Another alternative available in some software packages is creating linear combinations of the explanatory variables, then entering these as new explanatory variables for the tree.

Was this article helpful?

0 0
Solar Power

Solar Power

Start Saving On Your Electricity Bills Using The Power of the Sun And Other Natural Resources!

Get My Free Ebook


Post a comment