Table 25 Temporal consistency analysis across different training strategies.
Training strategy | Stable condition consistency (95% CI) | Progression detection accuracy (95% CI) | Regression detection accuracy (95% CI) | False change rate (95% CI) | Temporal reliability score (mean ± SD) | Cronbach’s α | p-value |
|---|---|---|---|---|---|---|---|
Real data only | 91.2% (89.8–92.6) | 78.4% (76.1–80.7) | 73.6% (71.0-76.2) | 8.8% (7.6–10.0) | 7.2 ± 0.6/10 | 0.823 | - |
25% synthetic | 92.7% (91.4–94.0) | 84.1% (82.1–86.1) | 79.8% (77.4–82.2) | 7.3% (6.2–8.4) | 8.1 ± 0.4/10 | 0.867 | < 0.001 |
50% synthetic | 93.9% (92.7–95.1) | 88.6% (86.9–90.3) | 84.2% (82.1–86.3) | 6.1% (5.1–7.1) | 8.7 ± 0.3/10 | 0.891 | < 0.001 |
75% synthetic (optimal) | 94.3% (93.2–95.4) | 91.2% (89.7–92.7) | 87.4% (85.5–89.3) | 5.7% (4.8–6.6) | 9.1 ± 0.2/10 | 0.912 | < 0.001 |
100% synthetic | 92.1% (90.7–93.5) | 87.9% (86.1–89.7) | 82.6% (80.4–84.8) | 7.9% (6.8-9.0) | 8.3 ± 0.4/10 | 0.876 | < 0.001 |
