A Note on Rescaling the Arithmetic Mean for Right-skewed Positive Distributions

Abstract: 

When the arithmetic mean (mean) is used as a measure of location for a set of right-skewed positive observations, it is subject to being pulled upward.  This upward movement tends to move the mean away from the bulk of the observations, making it less representative of them.  One way to deal with this loss of representativeness is to transform the data.  A Box-Cox power trans-formation can make a right-skewed distribution more symmetrical and then a measure of location for the original observations is found by applying an inverse transformation to the center of the transformed data.  This approach was used in a series of papers dealing with the Mean Absolute Percent Error (MAPE) as a measure of forecast and estimation error.  In this paper, we show that the Box-Cox power transformation can be used more generally with any mean computed for a set of right-skewed positive observations to develop R-MEAN (Rescaled-Mean).  We provide a set of examples to illustrate this approach and show its use in an actual application.

Keywords:  

Asymmetric distribution; Box-Cox Power Transformation; Outlier; R-MEAN

JEL Classification: 

B41, C13, C18

Citation as:  

Swanson, D. A., J. Tayman, and T.M. Bryan(2018). "A Note on Rescaling the Arithmetic Mean for Right-skewed Positive Distributions", Review of Economics & Finance, 14(4): 17-24.