Bioelectrical impedance: fat content of beef and pork from different size grinds

Multiple linear regression equations were developed for predicting the percentage of fat content of beef and pork. The predictor variables were bioelectrical resistance, temperature, and weight of product. Equations were developed for trim and product ground through a .95- or a .32-cm plate. The trim, .95-cm, and .32-cm grinds had 64, 108, and 96 observations, respectively, for beef product and 56, 101, and 92 observations, respectively, for pork product. Each of these observations was the average of bioelectrical impedance measurements taken in triplicate. The fat percentage ranges were 4 to 50% for beef and 7.5 to 50% for pork. The prediction equation applied to beef trim provided the following values: R(2) = .80, Mallows's Cp = 5.1, and root mean square error = 6.64. The R(2) for equations predicting fat percentage in .95- and .32-cm ground beef were .84 and .95, respectively. The prediction equation applied to pork trim provided the following values: R(2) = .77, Mallow's Cp = 5.0, and root mean square error = 6.2. The R(2) for equations predicting fat percentage in .95- and .32-cm ground pork were .87 and .96, respectively. The analyses were repeated with data sets of observations with less than 35% fat. The sample sizes and R(2) for the trim, .95-, and .32-cm ground beef were 48, .36; 76, .60; and 65, .86; respectively. The sample sizes and R(2) for the trim, .95-, and .32-cm ground pork were 42, .64; 62, .66; and 58, .92; respectively. Resistance, temperature, and weight remained as predictor variables for ground product with less than 35% fat. The smaller the grind, the more accurate the prediction. These results are positive for developing inexpensive, on-line systems for efficiently mixing ground product to a specific fat percentage.