Methods and formulas for the percent of error statistics in Random Forests® Regression


This command is available with the Predictive Analytics Module. Click here for more information about how to activate the module.

For the percent of error statistics, the value depends on the percentage of the largest residuals in the calculation. In the following formulas, the calculations assume that the residuals are in order by absolute value, such that i = 1 represents the residual with the greatest absolute value and i = N represents the residual with the least absolute value.

For out-of-bag validation, the fit for a row comes from the average of the predictions when the row is in the out-of-bag data.

When you use a test data set for validation, the test statistics use fitted values for the test data set only.





ccount of largest residuals for the percentage
yi observed response value
mean response
fitted response
Nnumber of rows