Table A1.
Comparison of logistic regression models.
| Model | p-Values of each coefficient | BIC | AIC |
|---|---|---|---|
| 1 + Age + roll 0.2 Hz (age adjusted) | 0.00017 0.0047 | 84.9 | 77.1 |
| 1 + First PCA component | 0.000075 | 85.2 | 80 |
| 1 + Age + roll 0.2 Hz | 0.013 0.0058 | 85.4 | 77.7 |
| 1 + Age + first PCA component (age adjusted) | 0.00032 0.0087 | 86.1 | 78.3 |
| 1 + Age + first PCA component | 0.098 0.013 | 86.9 | 79.1 |
| 1 + Roll 0.2 Hz | 0.000081 | 88 | 82.8 |
| 1 + Age + yaw + roll 0.2 Hz | 0.014 0.36 0.014 | 89.2 | 78.8 |
| 1 + Age + roll 1 Hz | 0.040 0.036 | 89.3 | 81.5 |
| 1 + Age + roll 1 Hz + roll 0.2 Hz | 0.063 0.42 0.039 | 89.4 | 79 |
| 1 + Age + sex + roll 0.2 Hz | 0.016 0.42 0.0055 | 89.4 | 79 |
| 1 + First PCA component + second PCA component | 0.000087 0.54 | 89.4 | 81.6 |
| 1 + Age + Z + roll 0.2 Hz | 0.063 0.52 0.018 | 89.6 | 79.3 |
| 1 + Age + Y + roll 0.2 Hz | 0.012 0.63 0.0072 | 89.8 | 79.4 |
| 1 + Age + first PCA component + second PCA component (age adjusted) | 0.00033 0.010 0.65 | 90.5 | 80.1 |
| 1 + Age + Z | 0.026 0.090 | 91.1 | 83.3 |
| 1 + Age + first PCA component + second PCA component | 0.12 0.013 0.77 | 91.4 | 81 |
| 1 + Age + yaw | 0.00060 0.11 | 91.4 | 83.6 |
| 1 + Age + Y | 0.0010 0.43 | 93.4 | 85.7 |
| 1 + Age + Y + roll 1 Hz + roll 0.2 Hz | 0.055 0.53 0.37 0.033 | 93.6 | 80.6 |
| 1 + First PCA component (age adjusted) | 0.0052 | 99.8 | 94.6 |
| 1 + Age + yaw + Y + Z + roll 1 Hz + roll 0.2 Hz | 0.084 0.41 0.41 0.75 0.52 0.044 | 102 | 83.6 |
| 1 + First PCA component + second PCA component (age adjusted) | 0.0058 0.63 | 104 | 96.4 |
| 1 + Age + sex + yaw + Y + Z + roll 1 Hz + roll 0.2 Hz (age adjusted) | 0.00026 0.54 0.45 0.46 0.67 0.47 0.047 | 105 | 84.3 |
| 1 + Age + sex + yaw + Y + Z + roll 1 Hz + roll 0.2 Hz | 0.10 0.59 0.48 0.49 0.77 0.50 0.046 | 106 | 85.3 |
Most models included an intercept term (indicated by 1+). Most analyses used log-transformed thresholds, and some used age-adjusted, log-transformed thresholds (indicated by age adjusted). The second column shows the p values for each term, excluding the intercept term. The third and fourth terms show the Bayesian information criterion (BIC) and Akaike information criterion (AIC) for each model, with a lower value indicating a better model.
PCA, principal component analysis.