Prediction of Marbling Scores From Percentage Intramuscular Fat

In this study two separate sets of data were used to develop prediction equations for Marbling Scores (MS) from actual percentage intramuscular fat. All regression parameters were significantly (P<.01) different from zero. Model-I explained only 49% of the variation in MS as compared with 83% for modelII. When these equations were validated on an independent data set, the correlations between the predicted and actual marbling scores were similar at 0.69 and 0.66 for model-I and model-II, respectively. Model-I has correctly classified 45.45%, 67.62%, 63.83%, and 42.86% of the bservations in trace, slight, small and modest classes, respectively. ModelII performed better than model-I for the extreme MS classes (practically devoid and moderate).