COMBINING MULTIPLE MARKERS FOR MULTI‐CATEGORY CLASSIFICATION: AN ROC SURFACE APPROACH

The Receiver Operating Characteristic (ROC) curve and the Area Under the ROC Curve (AUC) are effective statistical tools for evaluating the accuracy of diagnostic tests for binary‐class medical data. However, many real‐world biomedical problems involve more than two categories. The Volume Under the ROC Surface (VUS) and Hypervolume Under the ROC Manifold (HUM) measures are extensions for the AUC under three‐class and multiple‐class models. Inference methods for such measures have been proposed recently. We develop a method of constructing a linear combination of markers for which the VUS or HUM of the combined markers is maximized. Asymptotic validity of the estimator is justified by extending the results for maximum rank correlation estimation that are well known in econometrics. A bootstrap resampling method is then applied to estimate the sampling variability. Simulations and examples are provided to demonstrate our methods.     

Semiparametric ROC surface estimation for continuous diagnostic tests via polytomous logistic regression procedures

In this paper, we propose a semiparametric method of estimating receiver operating characteristic (ROC) surfaces for continuous diagnostic tests under density ratio models. Implementation of our method is easy since the usual polytomous logistic regression procedures in many statistical software packages can be employed. A simulated example is provided to facilitate the implementation of our method. Simulation results show that the proposed semiparametric ROC surface estimator is more efficient than the nonparametric counterpart and the parametric counterpart whether the normality assumption of data holds or not. Moreover, some simulation results on the underlying semiparametric distribution function estimators are also reported. In addition, some discussions on the proposed method as well as analysis of a real data set are provided.     

A modified area under the ROC curve and its application to marker selection and classification

The area under the ROC curve (AUC) can be interpreted as the probability that the classification scores of a diseased subject is larger than that of a non-diseased subject for a randomly sampled pair of subjects. From the perspective of classification, we want to find a way to separate two groups as distinctly as possible via AUC. When the difference of the scores of a marker is small, its impact on classification is less important. Thus, a new diagnostic/classification measure based on a modified area under the ROC curve (mAUC) is proposed, which is defined as a weighted sum of two AUCs, where the AUC with the smaller difference is assigned a lower weight, and vice versa. Using mAUC is robust in the sense that mAUC gets larger as AUC gets larger as long as they are not equal. Moreover, in many diagnostic situations, only a specific range of specificity is of interest. Under normal distributions, we show that if the AUCs of two markers are within similar ranges, the larger mAUC implies the larger partial AUC for a given specificity. This property of mAUC will help to identify the marker with the higher partial AUC, even when the AUCs are similar. Two nonparametric estimates of an mAUC and their variances are given. We also suggest the use of mAUC as the objective function for classification, and the use of the gradient Lasso algorithm for classifier construction and marker selection. Application to simulation datasets and real microarray gene expression datasets show that our method finds a linear classifier with a higher ROC curve than some other existing linear classifiers, especially in the range of low false positive rates.     

Estimation of the area under ROC curve with censored data

The area under the receiver operating characteristic curve is the most commonly used summary measure of diagnostic accuracy for a continuous-scale diagnostic test. In this paper, we develop methods to estimate the area under the curve (AUC) with censored data. Based on two different integration representations of this parameter, two nonparametric estimators are defined by the “plug in” method. Both the proposed estimators are shown to be asymptotically normal based on counting process and martingale theory. A simulation study is conducted to evaluate the performances of the proposed estimators.     

A Framework for Random-Effects ROC Analysis: Biases with the Bootstrap and Other Variance Estimators

In this article, we analyze the three-way bootstrap estimate of the variance of the reader-averaged nonparametric area under the receiver operating characteristic (ROC) curve. The setting for this work is medical imaging, and the experimental design involves sampling from three distributions: a set of normal and diseased cases (patients), and a set of readers (doctors). The experiment we consider is fully crossed in that each reader reads each case. A reading generates a score that indicates the reader's level of suspicion that the patient is diseased. The distribution of scores for the normal patients is compared to the distribution of scores for the diseased patients via an ROC curve, and the area under the ROC curve (AUC) summarizes the reader's diagnostic ability to separate the normal patients from the diseased ones. We find that the bootstrap estimate of the variance of the reader-averaged AUC is biased, and we represent this bias in terms of moments of success outcomes. This representation helps unify and improve several current methods for multi-reader multi-case (MRMC) ROC analysis.     

The average area under correlated receiver operating characteristic curves: a nonparametric approach based on generalized two-sample Wilcoxon statistics

It is well known that, when sample observations are independent, the area under the receiver operating characteristic (ROC) curve corresponds to the Wilcoxon statistics if the area is calculated by the trapezoidal rule. Correlated ROC curves arise often in medical research and have been studied by various parametric methods. On the basis of the Mann–Whitney U-statistics for clustered data proposed by Rosner and Grove, we construct an average ROC curve and derive nonparametric methods to estimate the area under the average curve for correlated ROC curves obtained from multiple readers. For the more complicated case where, in addition to multiple readers examining results on the same set of individuals, two or more diagnostic tests are involved, we derive analytic methods to compare the areas under correlated average ROC curves for these diagnostic tests. We demonstrate our methods in an example and compare our results with those obtained by other methods. The nonparametric average ROC curve and the analytic methods that we propose are easy to explain and simple to implement.     

Exact Bootstrap Variances of the Area Under ROC Curve

The area under the Receiver Operating Characteristic (ROC) curve (AUC) and related summary indices are widely used for assessment of accuracy of an individual and comparison of performances of several diagnostic systems in many areas including studies of human perception, decision making, and the regulatory approval process for new diagnostic technologies. Many investigators have suggested implementing the bootstrap approach to estimate variability of AUC-based indices. Corresponding bootstrap quantities are typically estimated by sampling a bootstrap distribution. Such a process, frequently termed Monte Carlo bootstrap, is often computationally burdensome and imposes an additional sampling error on the resulting estimates. In this article, we demonstrate that the exact or ideal (sampling error free) bootstrap variances of the nonparametric estimator of AUC can be computed directly, i.e., avoiding resampling of the original data, and we develop easy-to-use formulas to compute them. We derive the formulas for the variances of the AUC corresponding to a single given or random reader, and to the average over several given or randomly selected readers. The derived formulas provide an algorithm for computing the ideal bootstrap variances exactly and hence improve many bootstrap methods proposed earlier for analyzing AUCs by eliminating the sampling error and sometimes burdensome computations associated with a Monte Carlo (MC) approximation. In addition, the availability of closed-form solutions provides the potential for an analytical assessment of the properties of bootstrap variance estimators. Applications of the proposed method are shown on two experimentally ascertained datasets that illustrate settings commonly encountered in diagnostic imaging. In the context of the two examples we also demonstrate the magnitude of the effect of the sampling error of the MC estimators on the resulting inferences.     

An alternative method for global and partial comparison of two diagnostic systems based on ROC curves

In this paper, an alternative method for the comparison of two diagnostic systems based on receiver operating characteristic (ROC) curves is presented. ROC curve analysis is often used as a statistical tool for the evaluation of diagnostic systems. However, in general, the comparison of ROC curves is not straightforward, in particular, when they cross each other. A similar difficulty is also observed in the multi-objective optimization field where sets of solutions defining fronts must be compared with a multi-dimensional space. Thus, the proposed methodology is based on a procedure used to compare the performance of distinct multi-objective optimization algorithms. In general, methods based on the area under the ROC curves are not sensitive to the existence of crossing points between the curves. The new approach can deal with this situation and also allows the comparison of partial portions of ROC curves according to particular values of sensitivity and specificity of practical interest. Simulations results are presented. For illustration purposes, considering real data from newborns with very low birthweight, the new method was applied in order to discriminate the better index for evaluating the risk of death.     

Robustness of Approaches to ROC Curve Modeling under Misspecification of the Underlying Probability Model

A variety of statistical regression models have been proposed for the comparison of ROC curves for different markers across covariate groups. Pepe developed parametric models for the ROC curve that induce a semiparametric model for the market distributions to relax the strong assumptions in fully parametric models. We investigate the analysis of the power ROC curve using these ROC-GLM models compared to the parametric exponential model and the estimating equations derived from the usual partial likelihood methods in time-to-event analyses. In exploring the robustness to violations of distributional assumptions, we find that the ROC-GLM provides an extra measure of robustness.     

Simex approaches to measurement error in roc studies

This paper explores the estimation of the area under the ROC curve when test scores are subject to errors. The naive approach that ignores measurement errors generally yields inconsistent estimates. Finding the asymptotic bias of the naive estimator, Coffin and Sukhatme (1995, 1997) proposed bias-corrected estimators for parametric and nonparametric cases. However, the asymptotic distributions of these estimators have not been developed because of their complexity. We propose several alternative approaches, including the SIMEX procedure of Cook and Stefanski (1994). We also provide the asymptotic distributions of the SIMEX estimators for use in statistical inference. Small simulation studies illustrate that the SIMEX estimators perform reasonably well when compared to the bias-corrected estimators.     

CONFIDENCE INTERVALS FOR THE DIFFERENCE BETWEEN TWO PARTIAL AUCS

As new diagnostic tests are developed and marketed, it is very important to be able to compare the accuracy of a given two continuous‐scale diagnostic tests. An effective method to evaluate the difference between the diagnostic accuracy of two tests is to compare partial areas under the receiver operating characteristic curves (AUCs). In this paper, we review existing parametric methods. Then, we propose a new semiparametric method and a new nonparametric method to investigate the difference between two partial AUCs. For the difference between two partial AUCs under each method, we derive a normal approximation, define an empirical log‐likelihood ratio, and show that the empirical log‐likelihood ratio follows a scaled chi‐square distribution. We construct five confidence intervals for the difference based on normal approximation, bootstrap, and empirical likelihood methods. Finally, extensive simulation studies are conducted to compare the finite‐sample performances of these intervals, and a real example is used as an application of our recommended intervals. The simulation results indicate that the proposed hybrid bootstrap and empirical likelihood intervals outperform other existing intervals in most cases.     

Applications of the Bootstrap in ROC Analysis

The problem of estimating standard errors for diagnostic accuracy measures might be challenging for many complicated models. We can address such a problem by using the Bootstrap methods to blunt its technical edge with resampled empirical distributions. We consider two cases where bootstrap methods can successfully improve our knowledge of the sampling variability of the diagnostic accuracy estimators. The first application is to make inference for the area under the ROC curve resulted from a functional logistic regression model which is a sophisticated modelling device to describe the relationship between a dichotomous response and multiple covariates. We consider using this regression method to model the predictive effects of multiple independent variables on the occurrence of a disease. The accuracy measures, such as the area under the ROC curve (AUC) are developed from the functional regression. Asymptotical results for the empirical estimators are provided to facilitate inferences. The second application is to test the difference of two weighted areas under the ROC curve (WAUC) from a paired two sample study. The correlation between the two WAUC complicates the asymptotic distribution of the test statistic. We then employ the bootstrap methods to gain satisfactory inference results. Simulations and examples are supplied in this article to confirm the merits of the bootstrap methods.     

Compare diagnostic tests using transformation-invariant smoothed ROC curves()

Receiver operating characteristic (ROC) curve, plotting true positive rates against false positive rates as threshold varies, is an important tool for evaluating biomarkers in diagnostic medicine studies. By definition, ROC curve is monotone increasing from 0 to 1 and is invariant to any monotone transformation of test results. And it is often a curve with certain level of smoothness when test results from the diseased and non-diseased subjects follow continuous distributions. Most existing ROC curve estimation methods do not guarantee all of these properties. One of the exceptions is Du and Tang (2009) which applies certain monotone spline regression procedure to empirical ROC estimates. However, their method does not consider the inherent correlations between empirical ROC estimates. This makes the derivation of the asymptotic properties very difficult. In this paper we propose a penalized weighted least square estimation method, which incorporates the covariance between empirical ROC estimates as a weight matrix. The resulting estimator satisfies all the aforementioned properties, and we show that it is also consistent. Then a resampling approach is used to extend our method for comparisons of two or more diagnostic tests. Our simulations show a significantly improved performance over the existing method, especially for steep ROC curves. We then apply the proposed method to a cancer diagnostic study that compares several newly developed diagnostic biomarkers to a traditional one.     

Three-group ROC predictive analysis for ordinal outcomes

Measuring the accuracy of diagnostic tests is crucial in many application areas including medicine, machine learning, and credit scoring. The receiver operating characteristic (ROC) surface is a useful tool to assess the ability of a diagnostic test to discriminate among three-ordered classes or groups. In this article, nonparametric predictive inference (NPI) for three-group ROC analysis for ordinal outcomes is presented. NPI is a frequentist statistical method that is explicitly aimed at using few modeling assumptions, enabled through the use of lower and upper probabilities to quantify uncertainty. This article also includes results on the volumes under the ROC surfaces and consideration of the choice of decision thresholds for the diagnosis. Two examples are provided to illustrate our method.     

Nonparametric estimation of the ROC curve based on smoothed empirical distribution functions

The receiver operating characteristic (ROC) curve is a graphical representation of the relationship between false positive and true positive rates. It is a widely used statistical tool for describing the accuracy of a diagnostic test. In this paper we propose a new nonparametric ROC curve estimator based on the smoothed empirical distribution functions. We prove its strong consistency and perform a simulation study to compare it with some other popular nonparametric estimators of the ROC curve. We also apply the proposed method to a real data set.     

Estimation methods for time‐dependent AUC models with survival data

The performance of clinical tests for disease screening is often evaluated using the area under the receiver‐operating characteristic (ROC) curve (AUC). Recent developments have extended the traditional setting to the AUC with binary time‐varying failure status. Without considering covariates, our first theme is to propose a simple and easily computed nonparametric estimator for the time‐dependent AUC. Moreover, we use generalized linear models with time‐varying coefficients to characterize the time‐dependent AUC as a function of covariate values. The corresponding estimation procedures are proposed to estimate the parameter functions of interest. The derived limiting Gaussian processes and the estimated asymptotic variances enable us to construct the approximated confidence regions for the AUCs. The finite sample properties of our proposed estimators and inference procedures are examined through extensive simulations. An analysis of the AIDS Clinical Trials Group (ACTG) 175 data is further presented to show the applicability of the proposed methods. The Canadian Journal of Statistics 38:8–26; 2010 © 2009 Statistical Society of Canada     

Powerful nonparametric statistics to compare k independent ROC curves

The authors deal with the problem of comparing receiver operating characteristic (ROC) curves from independent samples. From a nonparametric approach, they propose and study three different statistics. Their asymptotic distributions are obtained and a resample plan is considered. In order to study the statistical power of the introduced statistics, a simulation study is carried out. The (observed) results suggest that, for the considered models, the new statistics are more powerful than the usually employed ones (the Venkatraman test and the usual area under the ROC curve criterion) in non-uniform dominance situations and quite good otherwise.     

Generalized Confidence Interval Estimation for the Difference in Paired Areas Under the ROC Curves in the Absence of a Gold Standard

Receiver operating characteristic (ROC) curves can be used to assess the accuracy of tests measured on ordinal or continuous scales. The most commonly used measure for the overall diagnostic accuracy of diagnostic tests is the area under the ROC curve (AUC). A gold standard (GS) test on the true disease status is required to estimate the AUC. However, a GS test may be too expensive or infeasible. In many medical researches, the true disease status of the subjects may remain unknown. Under the normality assumption on test results from each disease group of subjects, we propose a heuristic method of estimating confidence intervals for the difference in paired AUCs of two diagnostic tests in the absence of a GS reference. This heuristic method is a three-stage method by combining the expectation-maximization (EM) algorithm, bootstrap method, and an estimation based on asymptotic generalized pivotal quantities (GPQs) to construct generalized confidence intervals for the difference in paired AUCs in the absence of a GS. Simulation results show that the proposed interval estimation procedure yields satisfactory coverage probabilities and expected interval lengths. The numerical example using a published dataset illustrates the proposed method.     

Maximum Likelihood Estimations and EM Algorithms with Length-biased Data

Length-biased sampling has been well recognized in economics, industrial reliability, etiology applications, epidemiological, genetic and cancer screening studies. Length-biased right-censored data have a unique data structure different from traditional survival data. The nonparametric and semiparametric estimations and inference methods for traditional survival data are not directly applicable for length-biased right-censored data. We propose new expectation-maximization algorithms for estimations based on full likelihoods involving infinite dimensional parameters under three settings for length-biased data: estimating nonparametric distribution function, estimating nonparametric hazard function under an increasing failure rate constraint, and jointly estimating baseline hazards function and the covariate coefficients under the Cox proportional hazards model. Extensive empirical simulation studies show that the maximum likelihood estimators perform well with moderate sample sizes and lead to more efficient estimators compared to the estimating equation approaches. The proposed estimates are also more robust to various right-censoring mechanisms. We prove the strong consistency properties of the estimators, and establish the asymptotic normality of the semi-parametric maximum likelihood estimators under the Cox model using modern empirical processes theory. We apply the proposed methods to a prevalent cohort medical study. Supplemental materials are available online.     

Information borrowing methods for covariate‐adjusted ROC curve

In medical diagnostic testing problems, the covariate adjusted receiver operating characteristic (ROC) curves have been discussed recently for achieving the best separation between disease and control. Due to various restrictions such as cost, the availability of patients, and ethical issues quite frequently only limited information is available. As a result, we are unlikely to have a large enough overall sample size to support reliable direct estimations of ROCs for all the underlying covariates of interest. For example, some genetic factors are less commonly observable compared with others. To get an accurate covariate adjusted ROC estimation, novel statistical methods are needed to effectively utilize the limited information. Therefore, it is desirable to use indirect estimates that borrow strength by employing values of the variables of interest from neighbouring covariates. In this paper we discuss two semiparametric exponential tilting models, where the density functions from different covariate levels share a common baseline density, and the parameters in the exponential tilting component reflect the difference among the covariates. With the proposed models, the estimated covariate adjusted ROC is much smoother and more efficient than the nonparametric counterpart without borrowing information from neighbouring covariates. A simulation study and a real data application are reported. The Canadian Journal of Statistics 40: 569–587; 2012 © 2012 Statistical Society of Canada