Table 2.

Thresholds to differentiate the different PA categories for the four acceleration summary measures and resulting accuracy with accelerometry data sampled at 10 Hz.

	ENMONZ	MAD	AI	ROCAM
Sleep	Estimated by using a separate sleep detection algorithm and additional entries that are below the lowest threshold of sedentary activity for each of the acceleration summary measures
Sedentary PA	0 < x ≤ 0.032	0.001 < x ≤ 0.059	0.010 < x ≤ 5.308	0.06 < x ≤ 0.175
Light PA	0.032 < x ≤ 0.173	0.059 < x ≤ 0.242	5.308 < x ≤ 17.010	0.175 < x ≤ 0.400
Moderate PA	0.173 < x ≤ 0.382	0.242 < x ≤ 0.38	17.010 < x ≤2 3.628	0.400 < x ≤ 0.483
Vigorous PA	x > 0.382	x > 0.38	x > 23.628	x > 0.483
Accuracy (%)	78.8	79.6	78.4	80.8

PA stands for Physical Activity. For determining sleep, we used a slightly modified algorithm that we had previously proposed [18] (see text for details). Using the sleep detection algorithm and the PA thresholds to estimate the five categories leads to the computed accuracies reported herein. For all acceleration summary measures, the results are presented in gravitational units (g).