Smoothed empirical likelihood for the Youden index
Introduction
In the past few decades, the receiver operating characteristic (ROC) curve has become a standard statistical tool to evaluate the discriminatory ability of a diagnostic test to separate diseased subjects from non-diseased subjects. For a diagnostic test with continuous scale, sensitivity (true positive rate) and specificity (true negative rate) are inversely related, in the sense that the increase of the one is accompanied with the decrease of the other as the cutoff point moves along the real number line. The ROC curve is the plot of sensitivity against 1-specificity at all possible threshold points. For comprehensive reviews of ROC analysis, see Shapiro (1999), Zhou et al. (2009), Pepe (2004), and Zou et al. (2010).
Even though the area under the ROC curve (AUC) has been widely used for measuring the accuracy of a diagnostic test, the Youden index has its unique advantage in advising an optimal cutoff for the clinicians to make diagnosis. The Youden index, firstly proposed by Youden (1950), is defined as the maximum of the sum of sensitivity and specificity minus one. The cutoff point, where the maximum is achieved, provides an optimal threshold for the clinicians to use the diagnostic test for classification if equal weight is placed on sensitivity and specificity. The possible values of the Youden index range from 0 to 1 with 0 indicating no discriminatory ability and 1 indicating perfect diagnostic accuracy. Graphically, the Youden index is the maximum vertical distance between the ROC curve and the diagonal chance line.
A variety of approaches have been developed for the inference of the Youden index. Fluss et al. (2005) provided several estimators for the Youden index and the associated cutoff points based upon normal assumption, empirical distribution function or kernel smoothing. For interval estimation of the Youden index, Schisterman and Perkins (2007) developed parametric methods under either normal or gamma assumption. Lai et al. (2012) made some improvement by utilizing a generalized variable approach. Even though monotonic transformation, such as Box–Cox transformation, can be applied, these parametric interval estimators may still be not satisfactory given the departures from distribution assumptions, particularly when the diseased and non-diseased populations are not from the same family of distributions. Nonparametric interval estimators of the Youden index are mainly developed from bootstrap. Based upon the Youden index estimators derived from different methods, including estimators under normal assumption, estimators from delta method, and estimators derived from empirical distribution function and kernel density estimation, a range of commonly used bootstrap methods, such as percentile, normal approximation and bias correction and acceleration (BCa) adjustment, have been considered by Faraggi (2003), Fluss et al. (2005) and Schisterman and Perkins (2007), respectively. More recently, Zhou and Qin (2012) proposed an adjusted bootstrap procedure via an approximate method for interval estimation of a single proportion introduced by Agresti and Coull (1998). Via extensive simulation studies, the authors suggested that their modified bootstrap method was comparable to parametric methods when distribution assumption holds unless the Youden index is close to upper boundary () and their methods outperformed the previously developed bootstrap methods.
In this paper, we aim to propose a novel interval estimator of the Youden index via the empirical likelihood. Empirical likelihood, formally proposed by Owen, 1988, Owen, 1990, is an appealing nonparametric method with many desirable features such as automatic determination of the shape of the confidence regions by data, straightforward incorporation of side information and being Bartlett correctable in many cases; see Owen (2001) for a comprehensive review. Claeskens et al. (2003) derived empirical likelihood confidence regions for ROC curves over a certain range of specificity values. Molanes-López and Letón (2011) proposed an empirical likelihood approach for the Youden index and its associated optimal cutoff point from a quantile function point of view. As it is not ready to profile out the nuisance parameter, the authors had to propose a fairly complicated two-cycle bootstrap procedure and the resulting interval estimator seemed to be over-conservative as suggested by simulation studies. In this paper, we develop empirical likelihood based upon novel estimation equations using kernel smoothing methods.
The rest of the paper is organized as follows: In Section 2, the novel empirical likelihood method is introduced. We also establish the asymptotic properties of the empirical likelihood ratio statistic and discuss potential computation algorithms. In Section 3, we evaluate the empirical performance of our method through extensive simulation studies. We illustrate the proposed method in Section 4 via the application to a published data set. We draw conclusions and make discussions in Section 5.
Section snippets
Methodologies
Let and denote diagnostic biomarker values from the diseased (case) and non-diseased (control) populations with distribution functions and , respectively. Without loss of generality, we assume that is stochastically less than (); otherwise the proposed method is still applicable to the negative of the biomarker values.
Let be the difference between the two cumulative distribution functions at a certain point . The Youden index can be expressed as
Simulation study
In this section, the empirical performance of our empirical likelihood method is assessed by extensive simulation studies. We reran the simulation experiments published in Zhou and Qin (2012) as follows:
- (i)
and , where the variance is set to be 0.5, 1, 3 and 5. For each value of , the mean is chosen such that the Youden index is equal to 0.4, 0.6, 0.8 and 0.9.
- (ii)
and , where the shape parameter is set to be 1.5, 2, 2.5 and 3. For each value of
Examples
We illustrate our empirical likelihood method through a data set of prostate cancer patients from Miller et al. (1980). The data set, as shown in Table 6, consists of the acid phosphatase levels in blood serum of 53 prostate cancer patients: of them without nodal involvement and of them with nodal involvement. The data set was previously analyzed by Le (2006) and Zhou and Qin (2012). Neither normal nor gamma distribution was found to fit the data well even after Box–Cox
Conclusions and discussions
In this paper, we develop empirical likelihood for the Youden index via defining novel estimating equations and establish Wilks theorem for the empirical likelihood ratio statistics. Simulation studies suggest that for small to medium sample sizes, the empirical likelihood interval estimators calibrated by distribution are robust under different distribution models. As compared to the bootstrap procedures, our empirical likelihood methods are more computational efficient and often have
Acknowledgments
We thank the editor-in-chief and two reviewers for their careful reading of the original manuscript and for their constructive comments which significantly improve the paper. The research of Yichuan Zhao is partially supported by NSF Grants (DMS-1406163 and DMS-1613176) and a NSA Grant (H98230-12-1-0209).
References (23)
- et al.
Exact confidence interval estimation for the Youden index and its corresponding optimal cut-point
Comput. Statist. Data Anal.
(2012) - et al.
Approximate is better than exact for interval estimation of binomial proportions
Amer. Statist.
(1998) - et al.
Smoothed empirical likelihood confidence intervals for quantiles
Ann. Statist.
(1993) - et al.
Empirical likelihood confidence regions for comparison distributions and roc curves
Canad. J. Statist.
(2003) Adjusting receiver operating characteristic curves and related indices for covariates
Statistician
(2003)- et al.
Estimation of the Youden index and its associated cutoff point
Biom. J.
(2005) A solution for the most basic optimization problem associated with ROC curve
Stat. Methods Med. Res.
(2006)- et al.
Biostatistics Casebook
(1980) - et al.
Inference of the Youden index and associated threshold using empirical likelihood for quantiles
Stat. Med.
(2011) Empirical likelihood ratio confidences for single functional
Biometrika
(1988)
Empirical likelihood ratio confidence regions
Ann. Statist.
Cited by (13)
A smooth nonparametric approach to determining cut-points of a continuous scale
2019, Computational Statistics and Data AnalysisCitation Excerpt :Some intuitive approaches include considering arbitrary cut-points; specifying cut-points as sample quantiles (e.g. median) or according to clinicians’ experience (Altman et al., 1994); using cut-points that yield disease rates consistent with a known population disease prevalence, or the highest proportions of correct classification based on a gold standard (Altman, 1991; Mazumdar and Glassman, 2000). Another popular approach is to utilize the receiver operating characteristic (ROC) curve (Pepe, 2003) in conjunction with various accuracy measures, such as Youden index (Youden, 1950; Fluss et al., 2005; Zhou and Qin, 2005; Schisterman et al., 2008; Zhou and Qin, 2012; Lai et al., 2012; Wang et al., 2017), concordance probability methods (Liu, 2012) and the point closest-to-(0, 1) corner method (Perkins and Schisterman, 2006), among others. Dong et al.
A New Classifier for Imbalanced Data Based on a Generalized Density Ratio Model
2023, Communications in Mathematics and StatisticsA review of recent advances in empirical likelihood
2023, Wiley Interdisciplinary Reviews: Computational StatisticsSmoothed empirical likelihood for optimal cut point analysis
2023, Communications in Statistics - Theory and Methods