Viabilidade do uso da função discriminante de Fisher: comparação com a MANAVA

Abstract

The multivariate analysis methods allow the simultaneous study when several variable responses are obtained by plot. An option for the treatment of multivariate data is the transformation, using Fisher's linear discriminant function (FDF). After reduction of the p- dimensional to the unidimensional space, the univariate analysis of variance (ANOVA) is applied. The objectives of this paper were to evaluate the transformation efficiency of the multivariate data through the FDF and to compare the detection capacity of differences between treatments by the ANOVA of these data with the results obtained by means of the multivariate analysis of variance (MANOVA). Simulations were carried out to evaluate the acceptance rates of the null hypotheses for treatments, in four levels of correlations, equality of averages and variances for treatments and inequality between averages and variances. It was applied ANOVA, MANOVA and ANOVA of FDF to the values of these simulations. The results of the simulations indicate that FDF is a proper alternative for data evaluation.