Data transformations in salary surveys

This article intends to show some of the aspects involved when elaborating and processing data coming from salary surveys to produce retribution reports. We will see how after trying different alternative transformations it is possible to verify that, through a logarithmic transformation, the asymmetry and kurtosis of a model's variables can be improved (revenues and salary) as well as a relation between two variables converted in lineal -something that would have been impossible to achieve with the original data. This would make data analysis easier, because it allows the use of the regression model. We will study the relation between salary and company size, because this relationship is the best to predict the market salary that is earned by a specific jobholder. We give an example of application of a regression model for studying this relationship. © 2013 Colegio Oficial de Psicólogos de Madrid. All rights reserved.