Bias–corrected methods for estimating the receiver operating characteristic surface of continuous diagnostic tests

Duc, Khanh To; Chiogna, Monica; Adimari, Gianfranco

doi:10.1214/16-ejs1202

Cited by 11 publications

(13 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the following example, the package is employed to evaluate the accuracy of ovarian cancer-related tumor marker CA125 (cancer antigen 125), which is a highly glycosylated sialomucin that is expressed on epithelial cell surface, especially on ovarian cancer cells. The dataset, available in the package, is described in detail in [ 5 ]. In what follows, CA125 will be the diagnostic test of interest and CA153 and Age will be two auxiliary covariates.…”

Section: Resultsmentioning

confidence: 99%

“…Table 2 shows the values of t -stat as well as p -values for testing H 0 : VUS = 1/6 vs. H 1 : VUS > 1/6. Statistical interpretation of results in Table 1 is given in [ 5 ], Section 5.1. Being this an artificial example, i.e., an example in which missingness of the disease has been artificially created according to the MAR assumption, the above mentioned section explains how to evaluate usefulness of bias correction and why SPE and IPW can be considered good estimates in this case.…”

Section: Resultsmentioning

confidence: 99%

“…From a computational point of view, the demand of bcROCsurface is essentially associated to estimation of variances of VUS (see [ 5 ]). To establish the computation time of the functions asyVarVUS() and vus(), a simulation study is performed.…”

Section: Resultsmentioning

confidence: 99%

“…Two main functions are provided: ROCs() for constructing and plotting ROC surfaces, and vus() for estimating VUS values as well as obtaining confidence intervals. Estimation methods can be flexibly selected by the argument method, among 6 options, i.e., full if the full data is available; fi for the FI estimator, msi for the MSI estimator, ipw for the IPW estimator, spe for the SPE estimator and knn for the KNN estimator in presence of partial verification (see [ 5 , 6 ] for the definition of the estimators). To plot ROC surfaces and ellipsoid confidence regions, the function ROCs() employs the plotting functions of the rgl package.…”

Section: Methodsmentioning

confidence: 99%

“…It provides several new functions for bias-corrected ROC surface analysis. More precisely, it implements methods in To Duc et al [ 5 , 6 ], who proposed five bias-corrected estimators for ROC surface and VUS of a continuous diagnostic test, namely, full imputation (FI), mean score imputation (MSI), inverse probability weighting (IPW), semiparametric efficient (SPE) and K nearest-neighbor (KNN) estimators. These methods perform provided that the missing mechanism is MAR (missing at random).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

bcROCsurface: an R package for correcting verification bias in estimation of the ROC surface and its volume for continuous diagnostic tests

Duc

2017

BMC Bioinformatics

Self Cite

View full text Add to dashboard Cite

BackgroundReceiver operating characteristic (ROC) surface analysis is usually employed to assess the accuracy of a medical diagnostic test when there are three ordered disease status (e.g. non-diseased, intermediate, diseased). In practice, verification bias can occur due to missingness of the true disease status and can lead to a distorted conclusion on diagnostic accuracy. In such situations, bias–corrected inference tools are required.ResultsThis paper introduce an R package, named bcROCsurface, which provides utility functions for verification bias–corrected ROC surface analysis. The shiny web application of the correction for verification bias in estimation of the ROC surface analysis is also developed.Conclusion bcROCsurface may become an important tool for the statistical evaluation of three–class diagnostic markers in presence of verification bias. The R package, readme and example data are available on CRAN. The web interface enables users less familiar with R to evaluate the accuracy of diagnostic tests, and can be found at http://khanhtoduc.shinyapps.io/bcROCsurface_shiny/.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

bcROCsurface: an R package for correcting verification bias in estimation of the ROC surface and its volume for continuous diagnostic tests

Duc

2017

BMC Bioinformatics

Self Cite

View full text Add to dashboard Cite

show abstract

Bayesian nonparametric estimation of ROC surface under verification bias

Zhu

Ghosal

2019

Statistics in Medicine

View full text Add to dashboard Cite

The receiver operating characteristic (ROC) surface, as a generalization of the ROC curve, has been widely used to assess the accuracy of a diagnostic test for three categories. A common problem is verification bias, referring to the situation where not all subjects have their true classes verified. In this paper, we consider the problem of estimating the ROC surface under verification bias. We adopt a Bayesian nonparametric approach by directly modeling the underlying distributions of the three categories by Dirichlet process mixture priors. We propose a robust computing algorithm by only imposing a missing at random assumption for the verification process but no assumption on the distributions.The method can also accommodate covariates information in estimating the ROC surface, which can lead to a more comprehensive understanding of the diagnostic accuracy. It can be adapted and hugely simplified in the case where there is no verification bias, and very fast computation is possible through the Bayesian bootstrap process. The proposed method is compared with other commonly used methods by extensive simulations. We find that the proposed method generally outperforms other approaches. Applying the method to two real datasets, the key findings are as follows: (1) human epididymis protein 4 has a slightly better diagnosis ability compared to CA125 in discriminating healthy, early stage, and late stage patients of epithelial ovarian cancer. (2) Serum albumin has a prognostic ability in distinguishing different stages of hepatocellular carcinoma.

show abstract

Estimation of the volume under the ROC surface in presence of nonignorable verification bias

2019

View full text Add to dashboard Cite

Bias–corrected methods for estimating the receiver operating characteristic surface of continuous diagnostic tests

Cited by 11 publications

References 13 publications

bcROCsurface: an R package for correcting verification bias in estimation of the ROC surface and its volume for continuous diagnostic tests

bcROCsurface: an R package for correcting verification bias in estimation of the ROC surface and its volume for continuous diagnostic tests

Bayesian nonparametric estimation of ROC surface under verification bias

Estimation of the volume under the ROC surface in presence of nonignorable verification bias

Contact Info

Product

Resources

About