### INTRODUCTION

### MATERIALS AND METHODS

### Animals and traits

### Statistical analyses

^{2}) and the residual standard deviation (RSD), and the equations with the highest R

^{2}and the lowest RSD values were considered to have the best predictive ability.

### RESULTS

### Stepwise regressions for RC and %RC

^{2}value was not significant.

^{2}of Eq. 3 with three independent variables were estimated as 10.350 kg and 87.1%, respectively, and the values for Eq. 4 in which MAR was added were 10.342 kg and 87.1%, respectively. The RSD and R

^{2}between Eq. 3 and Eq. 4 showed little difference for the dependent variables of RC and %RC.

### Stepwise regressions for FAT and %FAT

^{2}value by 5.3%, while the further addition of EMA increased the R

^{2}value to 68.8%. The MAR was the last variable in the equation, but its contribution to the variation in FAT was insignificant and induced an increase of only 0.2% in the equation evaluation with three variables (Eq. 3).

^{2}by 9.9% in the equation (from 17.7% in Eq. 1 to 27.6% in Eq. 3). However, the addition of MAR showed an increase in the R

^{2}of only 0.6% for the equation fit (from 27.6% in Eq. 3 to 28.2% in Eq. 4).

### Stepwise regressions for BONE and %BONE

^{2}values in Eq. 3 and Eq. 4 for BONE were 53.3% and 53.4%, respectively, which corresponded with 27.5% and 27.6% for %BONE, respectively. The differences in R

^{2}and RSD between the final four-variable equation (Eq. 4) and the three-variable equation in the third step (Eq. 3) were not very significant in BONE and %BONE.

### Evaluation of the equation

^{2}and small RSD, and they were applied to the test dataset for the evaluation of the equations. Each dependent variable predicted from Eq. 3 and Eq. 4 were compared with an extra retail cut percentage from the official equation (ELP).

### DISCUSSION

^{2}values, there were no practical differences between Eq. 3 and Eq. 4, which indicates that the last variables in the equation rarely contributed to the predictive ability of the equation. The inclusion of MAR, as the last variable in Eq. 3, with CWT, BFT, and EMA, increased the R

^{2}value of the equation for %RC and %FAT by only 0.02% and 0.06%, respectively (Tables 2 and 3). The inclusion of EMA, as the last variable in the Eq. 3, with CWT, BFT, and MAR increased the R

^{2}value of the equation for both BONE and %BONE by only 0.01% (Table 4).

^{2}values by sizable amounts (2% to 4%). The low contribution of MAR in predicting %RC and %FAT in this study could be due to a low correlation between MAR and %RC and %FAT in Hanwoo steer carcasses. Koh et al. (2014) reported a small and positive phenotypic correlation between MAR and %RC and %FAT in Hanwoo steer data (r = +0.04 and +0.07, respectively), of which similar and positive correlation coefficients of +0.02 and +0.03 were estimated in the preliminary analyses of this study (data not shown).

^{2}values of the equations for %RC, %FAT, and %BONE measured in percentages units, the R

^{2}values corresponding to the RC, FAT, and BONE in kg units, respectively, were higher, of which trends had generally been shown in previous studies (Herring et al., 1994; Shackelford et al., 1995; Williams et al., 1997; Dikeman et al., 1998; Realini et al., 2001; Greiner et al., 2003b; Lee et al., 2005; Maeno et al., 2014). These results imply that the equations for RC or FAT constructed using carcass traits might be more accurate than the equations for %RC and %FAT. Further, CWT among the independent variables showed the strongest correlation in the equation with the highest R

^{2}value as the best single predictor. The correlation coefficients for RC, FAT, and BONE with CWT, which were obtained using the square root of the R

^{2}values of Eq. 1 for each weight variable, were 0.91, 0.77, and 0.69, respectively, in this study.

^{2}values in Eq. 3 for RC and %RC were 87.1% and 23.5%, respectively, in the present study, and these seem to concur with the values reported by Lee et al. (2005). However, the R

^{2}value of 23.5% for %RC is lower than the 54% reported by Choy et al. (2010). On the other hand, the R

^{2}values reported for equations for %RC with exotic beef carcasses ranged from 32.2% (Williams et al., 1997) to 75% (Cannell et al., 1999), which are generally higher than the R

^{2}values from Eq. 3 and Eq. 4 in the present study. The different R

^{2}values from the various studies might be due to a number of factors, including different cattle types, feeding management, fat trimming level, carcass fatness, variables for equation development, and cutting procedures. Another plausible reason, which might be exclusive to the study on commercial data, was the inconsistent fat trimming level due to the purchaser’s demand. In this study, the fat trimming of the retail cut was conducted within the 6 mm fat cover, but the fat trimming level could differ based on the purchaser’s demand.

^{2}values or the correlation coefficients of the predicted and observed values (Johnson and Rogers, 1977; Tedeschi, 2006). In this study, almost identical correlation coefficients were found for Eq. 3 and Eq. 4 regarding the predicted and observed values for all the dependent variables, which indicates that equal precision was achieved by Eq. 3 and Eq. 4. The small average difference in the absolute value for Eq. 3 implies that Eq. 3 is more accurate than Eq. 4 in predicting %RC, BONE, and %BONE; furthermore, the three variables in the current official equation for predicting RC were found to still be valid. However, compared to Eq. 3 and Eq. 4 for predicting %RC, a relatively high absolute value in the average difference for the ELP suggests that further study may be necessary to revise the official equation.