Table 2.
Binary logistic regression models to predict woody breast (WB) condition in broiler carcasses using image measurements.
| Prediction model1 | Data set2 | Fit details |
Parameter estimates and odds ratio (OR) |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gen. R2 | RMSE | MAD | MR (%) | TPR (%) | FPR (%) | AUC | Parameter | Estimate | Std. Error | Pr > χ2 | OR | ||
| Model 1 | Training | 0.65 | 0.26 | 0.13 | 9.05 | 63.21 | 3.44 | 0.95 | M1 | 1.39 | 0.30 | <0.0001 | 4.03 |
| Validation | 0.61 | 0.26 | 0.13 | 8.52 | 71.11 | 4.44 | 0.94 | M2 | −7.55 | 1.00 | <0.0001 | 5.24 × 10−4 | |
| M3 | 1.83 | 0.28 | <0.0001 | 6.23 | |||||||||
| Model 2 | Training | 0.62 | 0.27 | 0.14 | 9.84 | 60.38 | 3.82 | 0.95 | M1 | 1.68 | 0.29 | <0.0001 | 5.38 |
| Validation | 0.58 | 0.27 | 0.15 | 9.63 | 64.44 | 4.44 | 0.93 | M2 | −4.40 | 0.99 | <0.0001 | 1.23 × 10−2 | |
| M4 | 0.33 | 0.06 | <0.0001 | 1.38 | |||||||||
| Model 3 | Training | 0.62 | 0.26 | 0.14 | 10.00 | 60.38 | 4.01 | 0.95 | M1 | 1.76 | 0.28 | <0.0001 | 5.82 |
| Validation | 0.58 | 0.27 | 0.14 | 8.52 | 68.89 | 4.00 | 0.93 | M2 | −9.45 | 1.06 | <0.0001 | 7.91 × 10−5 | |
| M6 | 0.31 | 0.06 | <0.0001 | 1.37 | |||||||||
| Model 4 | Training | 0.60 | 0.27 | 0.14 | 10.00 | 56.60 | 3.24 | 0.94 | M9 | 8.86 | 1.30 | <0.0001 | 7.07 × 103 |
| Validation | 0.58 | 0.27 | 0.15 | 10.00 | 57.78 | 3.56 | 0.93 | M11 | 26.28 | 6.55 | <0.0001 | 2.60 × 1011 | |
| Model 5 | Training | 0.63 | 0.26 | 0.14 | 10.79 | 58.49 | 4.58 | 0.95 | M1 | 0.63 | 0.11 | <0.0001 | 1.87 |
| Validation | 0.56 | 0.27 | 0.14 | 9.26 | 66.67 | 4.44 | 0.92 | M9 | 11.10 | 1.22 | <0.0001 | 6.62 × 104 | |
| Model 6 | Training | 0.63 | 0.26 | 0.14 | 10.00 | 60.38 | 4.01 | 0.95 | M3 | 1.09 | 0.15 | <0.0001 | 2.99 |
| Validation | 0.61 | 0.26 | 0.14 | 8.52 | 73.33 | 4.89 | 0.94 | M11 | 48.08 | 6.31 | <0.0001 | 7.59 × 1020 | |
M1, M2, …, Mn are the independent predictor variables (image measurements), α is the intercept, and β1, β2, …, βn are the regression coefficients.
Abbreviations: AUC, area under the ROC (receiver operating characteristic) curve; FPR, false positive rate; MAD, mean absolute deviation; MR, misclassification rate; RMSE, root mean square error; TPR, true positive rate (sensitivity–ST)
Model 1 = Logit (p) = α + β1 M1 + β2 M2 + β3 M3. Model 2 = Logit (p) = α + β1 M1 + β2 M2 + β3 M4. Model 3 = Logit (p) = α + β1 M1 + β2 M2 + β3 M6. Model 4 = Logit (p) = α + β1 M9 + β2 M11. Model 5 = Logit (p) = α + β1 M1 + β2 M9. Model 6 = Logit (p) = α + β1 M3 + β2 M11.
Training {n = 630: No [WB scores 0.0 or 0.5 (n = 376), and 1.0 or 1.5 (n = 148)], and Yes [WB scores 2.0 or 2.5 (n = 76), and 3.0 (n = 30)]} and validation {n = 270: No [WB scores 0.0 or 0.5 (n = 160), and 1.0 or 1.5 (n = 65)], and Yes [WB scores 2.0 or 2.5 (n = 35), and 3.0 (n = 10)]}.