Resumo. Principal components and factor analysis were used to evaluate the applicability of the linear model IVI = a 1 A r + a 2 D r + a 3 F r for obtaining indexes of value of importance ( IVI ) for tree species, where A r is relative abundance, D r is the relative dominance and F r is the relative frequency, to compare with the standard univariate method that considers a 1 = a 2 = a 3 =1 . The data comprised a sample of 104.04 ha, located in Tapajós National Forest, Belterra municipality, state of Pará, covering of a set of 36 clusters of 4 ha each, Malta croos-shaped, with four subplots of 1ha each. To obtain the IVI’s by species within each cluster, each subplot was subdivided each into 5 sub-subplots of 0.2 ha each. The analysis revealed that the values obtained by multivariable differ from those obtained by the standard equation. It was also found that the values of univariate and multivariate indices vary significantly within the forest, and therefore are not recommended, as they usually are, to infer a single value for the whole forest area, because species considered of minor importance for the overall forest may be of great importance in particular subareas. The use of linear model with a 1 = a 2 = a 3 =1 is not suitable, because of the wide variation of the coefficients among species and their spatial distribution in the forest.