Skip to main content

Table 3 Variance of estimates provided by the MI, TOBIT and LOGIT approaches at various proportions of nondetects (entries are averages of 1000 repetitions)

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

 

Sample size of 200

Sample size of 400

Sample size of 1000

% NDs

MI

TOBIT

Logit

MI

TOBIT

Logit

MI

TOBIT

Logit

10%

0.0050

0.0044

0.0358

0.0022

0.0022

0.0164

0.0008

0.0009

0.0063

30%

0.0055

0.0050

0.0185

0.0022

0.0025

0.0089

0.0007

0.0010

0.0035

50%

0.0055

0.0061

0.0195

0.0021

0.0030

0.0095

0.0006

0.0012

0.0037

70%

0.0048

0.0092

0.0319

0.0018

0.0043

0.0147

0.0005

0.0017

0.0056