Estimation of crude fiber content of a feed from its ADF value where there is no laboratory service

Because of the cost and inaccessibility of laboratory facilities, animal feed formulation at the farm level, in many parts of Ethiopia, is based on feed database information. However, nowadays many laboratories are phasing out the Weende crude fiber (CF) method of analysis. The fiber content of feeds available in most feed databases (including the sub-Saharan Africa feeds composition database) are a result of detergent method analysis (NDF, ADF and lignin). However, CF is still used in poultry feed formulation and forage analysis for horses, in addition to the neutral detergent fiber (NDF) fraction for determining fiber in different countries. Since there is a statistically (P<0.01) difference between the CF and acid detergent fiber (ADF) value of a feed, ADF can’t be used directly in place of CF. Therefore, this work aims to formulate a regression equation that could roughly estimate the CF level of a feed from its NDF and ADF values. Considering the strong multicollinearity between NDF and ADF, this study developed separate models for ADF and NDF and compared them based on R² and Akaike Information Criterion (AIC), and the ADF-based model provided a better fit. The equations 0.79×ADF-0.46, 0.01+0.79×ADF, and 1.37+0.62×ADF have effectively predicted CF for cereal grains and beans, pulses and byproducts, and also oilseed meals and cakes, respectively. For grass forages, the equation 3.38+0.76×ADF, tested on 10 forages, showed potential but remains unreliable due to its R² value below 0.8. Finally, it is concluded that this approach provides a practical alternative for estimating CF where laboratory services or database information are unavailable. Keywords: Crude Fiber, Estimation, Feed database information, Prediction, Regression.