View article

Near infrared reflectance spectroscopy (NIRS) was used to predict the chemical constituents, digestibility and energy value of 164 experimental compound feeds for rabbits using harmonised methods in six European institutes. NIRS equations were developed by partial least square regression (PLSR) using two-thirds of the samples as a calibration set and the remaining samples as a validation set. Prediction was satisfactory for the following characteristics: dry matter (coefficient of determination in validation, R_v²=0.70; standard error of prediction, SEP=4.8gkg⁻¹), crude protein (R_v²=0.86; SEP=5.6gkg⁻¹DM), ether extract (R_v²=0.93; SEP=4.2gkg⁻¹ DM), ADF (R_v²=0.82; SEP=14gkg⁻¹ DM), starch (R_v²=0.90; SEP=16gkg⁻¹DM), DM digestibility (R_v²=0.79; SEP=0.019), gross energy digestibility (R_v²=0.81; SEP=0.019) and digestible energy (R_v²=0.77; SEP=0.39MJkg⁻¹DM). NIRS prediction was less accurate for …