Table 3. Variable contribution assessment based on Kick-One-Out analysis using the BIC for predicting M/T ratio ≥1 in patients undergoing wedge resection.
| Variables | Reference | BIC | ΔBIC | Interpretation |
|---|---|---|---|---|
| BMI (≥25 kg/m2) | <25 kg/m2 | 75.27 | −4.10 | Low contribution |
| Smoking history (never) | Ever | 76.53 | −2.84 | Low contribution |
| Tumor location (upper or middle lobe) | Lower lobe | 75.31 | −4.07 | Low contribution |
| Marking position (supine) | Prone | 76.56 | −2.81 | Low contribution |
| Lipiodol spot size | – | 75.61 | −3.76 | Low contribution |
| Lipiodol localization (deep) | Shallow | 75.55 | −3.83 | Low contribution |
| 3D-deviation† | – | 75.43 | −3.94 | Low contribution |
| Tumor diameter | – | 84.19 | +5.81 | Strong positive contribution |
| Tumor depth (≥25.6 mm) | <25.6 mm | 82.80 | +3.42 | Moderate positive contribution |
ΔBIC = BIC_model − BIC_reference; smaller ΔBIC indicates better model fit. Interpretation: ΔBIC <2, negligible contribution; 2–4, moderate positive contribution; >4, strong positive contribution. †, 3D-deviation: distance between lipiodol spot and lesion. 3D, three-dimensional; BIC, Bayesian Information Criterion; BMI, body mass index; M/T, margin-to-tumor.