Transforms used in Regression

Four types of transforms on factor-levels or responses are generally useful: Individual parameters for a regression may be Centered or Normalized; they may have a Linear or a Nonlinear Transform.

Transformations have no useful purpose for qualitative factors.

An empirical choice of a transformation may be made based on two considerations. An effective transform will reduce the significance of interaction and quadratic terms and will simultaneously reduce the range and improve the uniformity of the regression residuals when plotted against significant factors of the regression. The Variance Analysis Ratio is often used for selecting an empirical transform.

Learn more about the Regression tools in Six Sigma Demystified (2011, McGraw-Hill) by Paul Keller, in his online Regression short course (only $99), or his online Black Belt certification training course ($875).