Transformação de dados com excesso de zero em experimentos com culturas olerícolas

Abstract

In olericulture, the highest interest is aimed at the production of commercial fruits. These experiments usually evaluate variables such as phytomass and number of fruits. However, these variables present a peculiar characteristic related to their presence or absence at the moment of harvesting. This attribute is represented by the occurrence of zero values for these variables, generating some variability and affecting the data variation analysis. In this study, experiments with summer squash, sweet pepper and broccoli were analyzed using the Box-Cox transformation family. Since it is restricted for null values, an adaptation suggested by Yamamura (1999) was used. It was concluded that for each culture and each variable studied, there is a need for differentiated transformations. In this context, in the experiment using summer squash, the indicated transformation is the square root. For sweet pepper, the indicated transformation is the inverted square root and for broccoli, the indicated transformation is the logarithmic function. Despite the reduced variability observed in all experiments, the use of the method was not sufficient to make these variances homoscedastic.