The objective of this work was to validate the prediction equations for neutral detergent fiber indigestible fraction (iNDF) employed by the Cornell Net Carbohydrate and Protein System (Sniffen equation) and the National Research Council (Conrad's equation). A database with 114 observations taken from tropical grass, with lignin and indigestible neutral detergent fiber (iNDF) contents, was used. Comparison among predicted and observed values was accomplished by adjusting a linear regression model with predicted values as dependent variable and observed values as independent variable. The estimates supplied by Conrad's equation were different from observed values, with constant bias of +5.60 percentile points. Alteration of the exponent that composes this equation, from .667 to .85, controlled the bias, guaranteeing the similarity among predicted and observed values. Lack of fitness was observed on the Sniffen equation, which caused under and overestimation of the iNDF contents for lignin concentrations lower and higher than 10.4% of NDF, respectively. This behavior is supported by the absence of non-linear relationships in this equation, constituting its main element of deficiency. The bias free of Conrad equation with empirical correction enlarges the accuracy of estimation process, leading to the lower mean square of the prediction error.
Cornell system; indigestible neutral detergent fiber; lignin; NRC system
