摘要
This article deals with correlating two variables that have values that fall below the known limit of detection (LOD) of the measuring device;these values are known as non-detects (NDs). We use simulation to compare several methods for estimating the association between two such variables. The most commonly used method, simple substitution, consists of replacing each ND with some representative value such as LOD/2. Spearman’s correlation, in which all NDs are assumed to be tied at some value just smaller than the LOD, is also used. We evaluate each method under several scenarios, including small to moderate sample size, moderate to large censoring proportions, extr</span><span style="font-family:Verdana;">eme imbalance in censoring proportions, and non-bivariate nor</span><span style="font-family:Verdana;">mal (BVN) data. In this article, we focus on the coverage probability of 95% confidence intervals obtained using each method. Confidence intervals using a maximum likelihood approach based on the assumption of BVN data have acceptable performance under most scenarios, even with non-BVN data. Intervals based on Spearman’s coefficient also perform well under many conditions. The methods are illustrated using real data taken from the biomarker literature.
This article deals with correlating two variables that have values that fall below the known limit of detection (LOD) of the measuring device;these values are known as non-detects (NDs). We use simulation to compare several methods for estimating the association between two such variables. The most commonly used method, simple substitution, consists of replacing each ND with some representative value such as LOD/2. Spearman’s correlation, in which all NDs are assumed to be tied at some value just smaller than the LOD, is also used. We evaluate each method under several scenarios, including small to moderate sample size, moderate to large censoring proportions, extr</span><span style="font-family:Verdana;">eme imbalance in censoring proportions, and non-bivariate nor</span><span style="font-family:Verdana;">mal (BVN) data. In this article, we focus on the coverage probability of 95% confidence intervals obtained using each method. Confidence intervals using a maximum likelihood approach based on the assumption of BVN data have acceptable performance under most scenarios, even with non-BVN data. Intervals based on Spearman’s coefficient also perform well under many conditions. The methods are illustrated using real data taken from the biomarker literature.
作者
Courtney E. McCracken
Stephen W. Looney
Courtney E. McCracken;Stephen W. Looney(Center for Research and Evaluation, Kaiser Permanente Georgia, Atlanta, USA;Department of Population Health Sciences, Augusta University, Augusta, USA)
