Computing Single Parameter Transformations

Abstract

Four methods for computing the ‘best’ value for the single parameter of seven families of transformations are considered. The families are power, modulus, folded-power, Guerrero–Johnson, Aranda-Ordaz symmetric, Aranda-Ordaz asymmetric, and power-logit. The four methods are two which use constructed variables; maximum likelihood; and a robust (against outliers) method based on maximum likelihood. Tables of expressions used by the various methods for the different families of transformations are given.

Keywords:

• Transformations
• Estimation