TABLE III.
Performance comparison between four univariate probability models. The number indicates the total number of epochs that specific model achieves the lowest KS distance or the best goodness-of-fit among all epochs of 15 tested subjects. The second last column shows the improvement ratio (defined as the number of improved KS fits among the total number of epochs being tested in a specific condition) of the bivariate inverse gaussian (invGauss) model with respiration (RP), as compared with the univariate inverse Gaussian model. The last column shows the percentage of points that fall within the 95% confidence bounds among all of fits with the bivariate invGauss model.
| Probability Models | % points within | |||||
|---|---|---|---|---|---|---|
| condition | gamma | Gaussian | lognormal | invGauss | invGauss(+RP) | 95% conf. bounds |
| supine, control | 0 | 3 | 2 | 11 | 11/15 | 86.6% |
| supine, ATR | 0 | 2 | 2 | 3 | 4/7 | 52.2% |
| supine, PROP | 0 | 3 | 4 | 8 | 5/8 | 95.7% |
| supine, DB | 0 | 4 | 8 | 3 | 9/15 | 84.6% |
| upright, control | 0 | 4 | 1 | 10 | 12/15 | 84.6% |
| upright, ATR | 0 | 2 | 1 | 4 | 3/7 | 45.7% |
| upright, PROP | 0 | 4 | 3 | 8 | 7/8 | 93.0% |
| upright, DB | 0 | 3 | 7 | 5 | 7/15 | 64.5% |