Table 3.
k | n | External test set | Leaked internal test set | Not leaked internal test set |
---|---|---|---|---|
3 | 25 | 0.330 (0.080) | 0.253 (0.050)* | 0.340 (0.049) |
3 | 50 | 0.330 (0.052) | 0.269 (0.035)* | 0.335 (0.034) |
3 | 100 | 0.347 (0.037) | 0.299 (0.033)* | 0.347 (0.032) |
3 | 250 | 0.321 (0.023) | 0.288 (0.016)* | 0.320 (0.015) |
10 | 25 | 0.096 (0.026) | 0.054 (0.016)* | 0.100 (0.020) |
10 | 50 | 0.098 (0.019) | 0.072 (0.012)* | 0.108 (0.013) |
10 | 100 | 0.096 (0.013) | 0.071 (0.007)* | 0.097 (0.007) |
10 | 250 | 0.100 (0.008) | 0.085 (0.005)* | 0.102 (0.005) |
36 | 25 | 0.025 (0.007) | 0.010 (0.002)* | 0.027 (0.004) |
36 | 50 | 0.027 (0.005) | 0.014 (0.002)* | 0.027 (0.002) |
36 | 100 | 0.028 (0.004) | 0.018 (0.001)* | 0.029 (0.002) |
36 | 250 | 0.027 (0.002) | 0.021 (0.001)* | 0.028 (0.001) |
Average balanced accuracy (standard deviation) obtained in the external and internal test sets.
k is the number of single-center datasets, each one containing n participants. CT: cortical thickness.
*One-tailed paired t-test Bonferroni adjusted p-value < 10−9 for the comparison with the external test set average balanced accuracy.