Skip to main content
. 2021 Jul 26;16(7):e0254813. doi: 10.1371/journal.pone.0254813

Table 2. Descriptive statistics of errors incurred by each method to detect gait events.

(ms) Method Mean SD LLoA ULoA ICC Sensitivity (%) PPV (%)
FORELIMB
WALK Hoof-on Pastern M1 -5 11 -26 16 0.9981 100 99
M2 -4 14 -33 24 0.9953 99 99
Cannon M1 -5 21 -47 37 0.9905 100 97
M3 -30 40 -109 49 0.9679 100 93
M4 14 72 -127 155 0.9254 90 100
Hoof-off Pastern M1 3 16 -28 34 0.9957 100 99
M2 -7 27 -61 46 0.9881 99 99
Cannon M1 1 19 -36 38 0.9939 100 97
M3 -15 65 -142 112 0.9287 100 93
M4 -35 117 -265 194 0.8456 90 100
TROT Hoof-on Pastern M1 -4*,^ 10 -24 15 0.9912 99 99
M2 -2 *, 9 -19 15 0.9938 99 99
Cannon M1 -9^, 23 -55 37 0.9515 99 100
M3 -26 21 -67 16 0.9567 100 94
M4 43 67 -89 175 0.8336 81 100
Hoof-off Pastern M1 -18* 23 -62 27 0.9581 99 99
M2 4*, 34 -63 71 0.9373 99 99
Cannon M1 -15 33 -80 50 0.9341 99 100
M3 62 114 -231 218 0.9559 100 94
M4 77 102 -123 277 0.7366 81 100
HINDLIMB
WALK Hoof-on Pastern M1 2*,^ 10 -19 22 0.9974 100 100
M2 -1*, 10 -22 19 0.9979 99 100
Cannon M1 -5^, 11 -26 17 0.9977 100 99
M3 -31 55 -139 77 0.9456 100 97
M4 161 412 -652 974 0.7013 90 100
Hoof-off Pastern M1 6^ 14 -22 34 0.9934 100 100
M2 -3 38 -77 72 0.9603 99 100
Cannon M1 -1^, 15 -29 28 0.9940 100 99
M3 -57 69 -292 77 0.9284 100 97
M4 -15 180 -368 338 0.6999 90 100
TROT Hoof-on Pastern M1 1*,^ 12 -23 26 0.9947 99 100
M2 -1*, 19 -37 36 0.9880 99 100
Cannon M1 -3^, 16 -34 29 0.9858 96 100
M3 -4 27 -56 48 0.9630 100 97
M4 98 147 -190 387 0.5439 90 100
Hoof-off Pastern M1 2*,^ 9 -16 20 0.9967 99 100
M2 15*, 21 -26 57 0.9773 99 100
Cannon M1 -7^, 14 -35 22 0.9881 96 100
M3 -144 62 -264 -23 0.9128 100 97
M4 -72 100 -267 124 0.6402 90 100

Descriptive statistics of the errors (ms) incurred by each method in detecting gait events and agreement with Mref for calculating stride durations; superscripts *,^ and indicate where the difference between two mean errors was statistically significant (p<0.01), as revealed by post hoc tests; sensitivity and positive predictive values (PPV) are given as percentages of the total number of events detected; for each gait event, the best performing method(s) is emboldened.