Evaluation of regression methods when immunological measurements are constrained by detection limits

BMC Immunology

Table 4 Application to real data: for the LOGIT model two cut-off points were used, median for Cytokine 1 and DL for Cytokine 2

	Methods	$\hat{β}$ (Slope)	SE( $\hat{β}$ )	$\hat{β}$ SE( $\hat{β}$ )	p-value
Cytokine 1	Standard methods
	Deletion	0.012	0.042	0.277	0.782
	Substitution of 0	-0.255	0.055	-4.645	< 0.0001
	Substitution of DL/2	-0.186	0.047	-3.951	0.0001
	Substitution of DL	-0.117	0.040	-2.879	0.005
	LOGIT (Median)	-0.198	0.139	-1.423	0.155
	Advanced methods
	ROS	-0.085	0.039	-2.184	0.030
	TOBIT	-0.190	0.048	-3.960	0.00008
	MI	-0.149	0.046	-3.264	0.0006
Cytokine 2	Standard methods
	Deletion	0.801	0.129	6.193	< 0.0001
	Substitution of 0	0.585	0.1101	5.311	< 0.0001
	Substitution of DL/2	0.545	0.093	5.842	< 0.0001
	Substitution of DL	0.504	0.079	6.418	< 0.0001
	LOGIT (DL)	0.208	0.113	1.841	0.066
	Advanced methods
	ROS	0.497	0.081	6.135	< 0.0001
	TOBIT	0.792	0.179	4.436	< 0.0001
	MI	0.550	0.127	4.322	< 0.0001

ISSN: 1471-2172