The interest in horticulture is the production of marketable fruit. In such experiments , phytomass and the number of fruits are commonly measured variables. However, these variables have a particular characteristic related to their absence or presence at harvest time. This attribute is represented by zero values for these variables, generating some variability and affecting the variance analysis. This work aimed to define an appropriate data transformation for the variables observed in experiments with the zucchini culture, realized in protected environment, designed to stabilize the variability values generated by the presence of zeros in harvest multiple fruit. The analysis was based on experiments with the zucchini culture ,using the family of the Box-Cox transformation, with an adaptation made by YAMAMURA (1999). It was concluded that for experiments with zucchini, assessing the phytomass and number of fruits, the square square root transformation is indicated. The same was observed with the reduction in variability and normal errors, observed in all experiments. The use of the method was not efficient to make the variances homocedastic.
Box-Cox; variance analysis; coefficient variation; heterocedastic variances
