. 2014 Nov;21(11):1512–1520. doi: 10.1128/CVI.00469-14

TABLE 3.

Summary of parsimonious variable selections^b

Variable name	Time point (mo)	Simple logistic regression results			BBR LASSO	Elasticnet		Glmnet
Variable name	Time point (mo)	Intercept (P value)	Parameter (P value)	AUC (95% Cl)	BBR LASSO	LASSO	Elastic net	LASSO	Elastic net
IgG	6	0.8952 (<0.0001)	1.7332 (<0.0001)^a	0.7724 (0.6956–0.8491)					X
	7	−0.6569 (0.0582)	1.0009 (<0.0001)^a	0.7956 (0.7208–0.8703)			X		X
	Last	0.7105 (0.0015)	2.1628 (<0.0001)^a	0.8214 (0.7514–0.8914)	•	•	•	•	•
TNA	6	−1.346 (0.0060)	1.8551 (<0.0001)^a	0.7416 (0.6689–0.8143)					X
	7	−1.9889 (0.0006)	1.0808 (<0.0001)^a	0.7918 (0.7158–0.8678)			X		X
SI	2	−4.9416 (<0.0001)	1.4900 (<0.0001)^a	0.7860 (0.7086–0.8633)	•	•	•	•	•
	6	−3.9370 (0.0376)	1.4369 (0.0111)^a	0.7095 (0.5809–0.8381)	•	•	•	•	•
IFNeLi	6	−0.3719 (0.3090)	0.7881 (0.0004)^a	0.7073 (0.6077–0.8068)					X
R_IL4IFNeLi	7	0.6808 (0.0020)	−0.1508 (0.3684)	0.5428 (0.4195–0.6661)					X
IL4e	1	−2.0654 (0.5856)	2.0972 (0.4594)	0.5149 (0.4790–0.5508)					X
TNFe	1	3.2168 (0.0787)	−1.4247 (0.1622)	0.6627 (0.5485–0.7769)					X
R_IL4IFNm	7	0.6608 (0.0004)	−0.7249 (0.0508)	0.5829 (0.4844–0.6815)					X
No. of variables					3	3	5	3	12
Variable set identifier					Par_LASSO^c	Par_LASSO^c	Par_elastic_Elasticnet	Par_LASSO^c	Par_elastic_Glmnet

P value of <0.05 from Wald chi-square test of the parameter estimate.

•, variables that were selected by all five selection methods. X, variables that were selected by 1 or 2 methods.

Parsimonious selections by LASSO in all the packages selected the same variables that were considered one variable set, Par_LASSO, for subsequent analyses.