Abstract
This study aimed to model nitrogen mineralization in an Oxisol using six nonlinear regression models and to evaluate their inferences by the curvature measures. Data from Pereira et al. (2005) were used, and simulations were conducted in R software. RMS curvatures were analyzed for each model, allowing the quality of the inferences to be assessed under small-sample conditions. The model proposed by Inubushi et al. (1985) showed the highest curvature, impairing inference, although simulations indicated improvements with at least eight replicates. It is concluded that curvature is a useful metric to guide the selection of nonlinear models.