A Modified Family of Power Transformations

• Main Author: YANG, Zhenlin
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2002
• Subjects: Box-Cox power transformation; Dual power transformation; Transformed regression model; Trans-normal distribution; Econometrics; Economics
• Online Access: https://ink.library.smu.edu.sg/soe_research/698; https://ink.library.smu.edu.sg/context/soe_research/article/1697/viewcontent/MTrans.pdf
• Institution: Singapore Management University

A modified family of power transformation, called the Dual Power Transformation, is proposed, which overcomes the truncation problem of the Box-Cox power transformation. The new transformation possesses properties similar to those of the Box-Cox power transformation. It generates a rich family of distributions that is seen to be very useful in modeling and analysis of economic durations and medical/engineering event-times. Further, it gives rise to transformed (regression) models such that all the standard asymptotic results of the maximum likelihood theory apply. Empirical results presented are more favorable to the new transformation than to the Box-Cox power transformation in terms of model fit.