The objective of this experiment was to use n-alkane to estimate accurately and precisely the botanical composition of dietary forage in sheep fed different proportions of Arachis pintoi Koprov & Gregory cv. Amarillo and Brachiaria decumbens Stapf (0, 15, 30, 45 e 60% of Arachis pintoi). Twenty lambs were used in a completely randomized design with 10 days for diet adaptation and 5 days for fecal collection. Both fecal composites and hay samples were analyzed for n-alkane concentration. Dietary forage proportions were estimated from measured concentrations of n-alkanes in feces and forage by the least square sum of deviations. Multivariate analyses and canonical variables were used to select the most discriminatory alkane with the obtained data submitted to analysis of variance. Means were compared using the "t" test and the adjustments of the estimated values in relation to the observed were done by linear regression. The canonical variables indicated that the alkanes C35, C33, C30, C31, C27, C29 and C36 had the greatest discriminatory potential. Therefore, the use of all these alkanes in the calculation was more accurate and precise (P <0.05) than the use of a combination of 2 or 3 alkanes to estimate the proportion of A. pintoi in the diet. The best adjustment of the regression was also found for these alkanes. The "t" test for the intercept of the equation (a) and the regression coefficient (b) showed that a = 0 and b = 1, indicating that the estimated values are equivalent to the observed. The multivariate analysis appears to be an important tool to select n-alkanes for diet evaluation.
cuticular wax; feed selection; herbage mixtures pasture; markers; multivariate analysis
