Table 2.
Model accuracies.
| Task | AUC | p-value | Spearman’s rho | p-value |
|---|---|---|---|---|
| SDMT | 0.73 | 0.0023 | 0.41 | 0.00005 |
| Cancelation | 0.73 | 0.002 | 0.27 | 0.005 |
| Paced tap | 0.70 | 0.008 | 0.05 | 0.2977 |
| Grip variability | 0.61 | 0.07 | 0.07 | 0.26453 |
| Stroop | 0.59 | 0.136 | 0.17 | 0.04877 |
| Count backwards | 0.54 | 0.314 | 0.08 | 0.19927 |
| Spot change | 0.52 | 0.396 | 0.01 | 0.4807 |
| Mental rotation | 0.51 | 0.439 | −0.01 | 0.54327 |
| Map search | 0.47 | 0.665 | −0.01 | 0.52993 |
| Indirect circle trace | 0.36 | 0.97 | −0.14 | 0.93302 |
Summary of AUC for classifications between fast-declining and stable subjects (extreme subgroups, subject numbers in Supplementary Table S5) using single tasks (permutation test). The Spearman rank correlation corresponds to the distance between the decision hyperplane and cognitive slopes using all pre-HD subjects from the test sites (within-site permutation test, see Methods).