Incorrect Use of Classic Bifactorial ANOVA: A Non-Parametric Alternative To Obtain Reliable Conclusions

ALLASIA MARÍA BELÉN; BOGGIONE MARÍA JULIA; SILVA QUINTANA S.; IVANCOVICH J.; PISKULIC L.; FERREYRA J.; LABADIÉ N.; BOTTAI H; RACCA L

BIOCELL

Instituto de Histología y Embriología; Centro Argentino de Información Científica y Tecnológica; Universidad Nacional de Cuyo

Lugar: Mendoza; Año: 2016 vol. 40 p. 6 - 6

1667-5746

Factorial experiments are widely used because of their efficiency to detect the effects of factors and their interactions. The statistical approach used in this context is Analysis of Variance (ANOVA). This classic method requires the fulfillment of some assumptions (normality, homoscedasticity and independence) which frequently are not met in practice. In these situations, nonparametric and robust methods are alternative approaches. The objective of this work is to show that a parametric approach can lead to erroneous conclusions when the assumptions are not fulfilled. We studied the effect of two factors: Support (soybean, paper) and Time (0,24,48 h) on fungal culture extract pH in cellulase production, under a completely randomized design. Classic ANOVA and Aligned Rank Transformation (ART, a non-parametric method) were performed on a simulated dataset. Normality was not verified (Anderson-Darling test, p