Table 3.

Logistic regression to discriminate healthy from MS.

Feat. No.	Sensor	Direction	Time	Arc	Calculation	Beta	S.E.	t	p
3	Thigh	Pitch	SI-ST	Arc F	AUC	1.0666e-03	3.34579e-04	3.188	0.0026
14	Spine	Pitch	SI-ST	Peak 2	Abs	0.17734	0.05628	3.151	0.0020
B	Stopwatch		Complete	TUG	Duration	−1.2315	0.3996	−3.082	0.0090
22	Spine	Roll	SI-ST	Arc D	Smoothness 2	−92.326	33.041	−2.794	0.0093
26	Spine	Roll	ST-SI	Peak 2	Abs	−0.10683	0.03984	−2.681	0.0103
17	Spine	Pitch	Turn 2	Arc Q	AUC	−9.3272e-04	4.4917e-04	−2.077	0.0086
27	Spine	Roll	ST-SI	Arc N	Smoothness 1	46.015	22.486	2.046	0.0209
	Constant					−10.149	5.322	−1.907	0.0099

For betas and t values that are positive, an increase in the feature's value implies the volunteer is healthy, while for negative values t and beta, higher values of the feature suggest that the volunteer is a person with MS. Rows are ordered by absolute value of t statistic, with the most contributory (and consistently discriminatory) features at the top. SI-ST, sit-to-stand transition; ST-SI, stand-to-sit transition; AUC, area under the curve; Abs, absolute value.