Table 4.
Comparison of peak linear acceleration (PLA) across six data cleaning methods
| Peak linear acceleration (PLA, g) | |||||
|---|---|---|---|---|---|
| Method | Median | 25th percentile | 75th percentile | Compared method | Corrected p-value |
|
1 (Uncleaned) |
14.6 | 13.6 | 16.7 | 2 | 1.000 |
| 3 | 1.000 | ||||
| 4 | 1.000 | ||||
| 5 | 1.000 | ||||
| 6 | 1.000 | ||||
|
2 (time-windowed HAEs only) |
15.0 | 13.5 | 16.0 | 3 | 1.000 |
| 4 | 1.000 | ||||
| 5 | 1.000 | ||||
| 6 | 1.000 | ||||
|
3 (algorithm-classified HAEs without time windowing) |
13.9 | 12.7 | 16.7 | 4 | 1.000 |
| 5 | 1.000 | ||||
| 6 | 1.000 | ||||
|
4 (algorithm-classified HAEs with time windowing) |
13.9 | 12.6 | 15.9 | 5 | 1.000 |
| 6 | 1.000 | ||||
|
5 (video verification) |
14.3 | 12.8 | 17.1 | 6 | 1.000 |
|
6 (HAEs confirmed by both proprietary classification algorithm and video verification) |
14.4 | 11.9 | 15.9 | ||
Median peak linear acceleration (PLA) values are presented for each data cleaning method. Differences in PLA across methods were assessed using a related-samples Friedman’s Two-Way ANOVA with significance values adjusted by a Bonferroni correction