Table 2. Logistic regression results using variables selected by adaptive Lasso (whole sample).

	Coefficient	OR	95% CI	P
Intercept	−69.203	–	–	0.014
Age of attaining the highest level	0.304	1.355	[1.052–1.789]	0.024
Gender, male	5.301	200.591	[32.352–1,577.368]	<0.0001
Ratio of arm to leg length A	0.190	1.209	[0.942–1.566]	0.141
Ratio of hand breadth to length	−0.153	0.859	[0.683–1.064]	0.174
Torso length	0.294	1.342	[1.148–1.585]	<0.0001
Leg length A to H	0.942	2.566	[1.601–4.313]	<0.0001
Ratio of bitrochanteric to biacromial breadth	−0.201	0.818	[0.698–0.947]	0.009
Head circumference	−0.787	0.455	[0.32–0.624]	<0.0001
Ratio of waist to chest	−0.264	0.768	[0.647–0.897]	0.001
Ratio of waist to hip	0.239	1.27	[1.107–1.474]	0.001
Biceps circumference difference	0.407	1.502	[0.86–2.692]	0.159
Skeletal muscle mass	0.287	1.332	[1.165–1.542]	<0.0001

Notes:

The penalty parameter was then chosen to shrink the coefficients of some of the candidate variables to zero (i.e., excluded from the model), and finally 12 out of the 24 variables were selected for the whole sample.

OR, odds ratio; CI, confidence interval.

Arm means upper limb length (cm); Leg length A means iliospinale posterior height (cm); Torso length means acromion height—trochanterion height (cm); Leg length H means eristailiaca height (cm); Biceps circumference difference means maximum biceps circumference—minimum biceps circumference (cm).