| Abstract: | Prediction of animal response from near infrared reflectance spectra of feeds was compared with predictions from chemical analyses. Sixty samples of pure and mixed forage-based diets were obtained from sheep intake and digestion trials. Sheep responses measured were digestible energy, dry matter intake, and calculated intake of digestible energy. Diets were analyzed chemically for protein, neutral detergent fiber, and in vitro dry matter disappearance. Coefficients of multiple determination and standard errors for fitting the sheep responses to these 60 diverse diets by regression equations developed from chemical analyses (.62 to .70) or spectra (.63 to .72) were similar. The 60 diets were divided into two sets of 30; one set was used to develop calibration equations for each sheep response, and the second set was used to test the equations. Calibration and errors of prediction were similar. When wavelengths chosen for each of the laboratory measurements were used to fit the sheep responses, standard errors were higher than when responses of sheep were predicted directly from spectra. The scanning instrument has the capability of predicting laboratory analyses and shows potential for predicting animal response as accurately as animal response can be predicted from laboratory analyses. |
