A Smooth ROC Curve Estimator Based on Log-Concave Density Estimates

Rufibach, Kaspar

doi:10.1515/1557-4679.1378

Cited by 7 publications

(11 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is well known that the kernel estimation of a density function requires a bandwidth of order N −1/5 , whereas the kernel estimation of the cumulative distribution function should be based on bandwidths of order N −1/3 . In the case of the ordinary ROC curve, smooth estimators have been proposed by Zou et al [21] and Lloyd [22], among others (see Rufibach [23] for a recent review). These estimators basically consist in the combination of kernel estimators of a cumulative distribution function and a quantile function, and require two smoothing parameters, one for each population.…”

Section: Smoothing Parameter Selectionmentioning

confidence: 99%

Smooth time-dependent receiver operating characteristic curve estimators

Martínez–Camblor

Pardo–Fernández

2017

Stat Methods Med Res

View full text Add to dashboard Cite

The receiver operating characteristic curve is a popular graphical method often used to study the diagnostic capacity of continuous (bio)markers. When the considered outcome is a time-dependent variable, two main extensions have been proposed: the cumulative/dynamic receiver operating characteristic curve and the incident/dynamic receiver operating characteristic curve. In both cases, the main problem for developing appropriate estimators is the estimation of the joint distribution of the variables time-to-event and marker. As usual, different approximations lead to different estimators. In this article, the authors explore the use of a bivariate kernel density estimator which accounts for censored observations in the sample and produces smooth estimators of the time-dependent receiver operating characteristic curves. The performance of the resulting cumulative/dynamic and incident/dynamic receiver operating characteristic curves is studied by means of Monte Carlo simulations. Additionally, the influence of the choice of the required smoothing parameters is explored. Finally, two real-applications are considered. An R package is also provided as a complement to this article.

show abstract

Section: Smoothing Parameter Selectionmentioning

confidence: 99%

Smooth time-dependent receiver operating characteristic curve estimators

Martínez–Camblor

Pardo–Fernández

2017

Stat Methods Med Res

View full text Add to dashboard Cite

show abstract

“…Figure 3 shows results obtained for 100,000 draws from the distributions specified earlier. The histograms of the three variances, shown in Figure 3(a) (note that only the lower 90% of simulated variances is included), show that the bulk of the probability mass ranges now from 0 to 8 5 . The histograms for the correlation coefficients, shown in Figure 3(b), show that for the first two components, the probability mass is equally spread between −1 and 1, in contrast to what can be observed in Figure 2(b).…”

Section: Lower 90% Of Simulated Covariance Component Sigma22mentioning

confidence: 99%

“…To show the impact of ignoring the imperfectness of the reference test, the logistic-regression model relating the clinical diagnosis to the three CSF biomarkers was additionally considered [39]. The AUC was computed based on log-condense smoothing of the empirical ROC curve as described by Rufibach [5] and FLAT, the analysis using the flat prior for sensitivity and specificity of the reference test; INF, the analysis using the informative prior for sensitivity and specificity of the reference test ( Figure 1a).…”

Section: Real Datamentioning

confidence: 99%

“…To show the impact of ignoring the imperfectness of the reference test, the logistic-regression model relating the clinical diagnosis to the three CSF biomarkers was additionally considered [39]. The AUC was computed based on log-condense smoothing of the empirical ROC curve as described by Rufibach [5] and implemented in the r-package pROC [40]. The resulting AUC distribution was obtained by bootstrapping.…”

Section: Datamentioning

confidence: 99%

“…The disadvantage of these methods is that the obtained ROC curves are usually not smooth. For this reason, non‐parametric ROC curve methods involving kernel density functions that result in smooth ROC curves were proposed .…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Development of a diagnostic test based on multiple continuous biomarkers with an imperfect reference test

Barrado

Coart

Burzykowski

2015

Statistics in Medicine

View full text Add to dashboard Cite

Ignoring the fact that the reference test used to establish the discriminative properties of a combination of diagnostic biomarkers is imperfect can lead to a biased estimate of the diagnostic accuracy of the combination. In this paper, we propose a Bayesian latent-class mixture model to select a combination of biomarkers that maximizes the area under the ROC curve (AUC), while taking into account the imperfect nature of the reference test. In particular, a method for specification of the prior for the mixture component parameters is developed that allows controlling the amount of prior information provided for the AUC. The properties of the model are evaluated by using a simulation study and an application to real data from Alzheimer's disease research. In the simulation study, 100 data sets are simulated for sample sizes ranging from 100 to 600 observations, with a varying correlation between biomarkers. The inclusion of an informative as well as a flat prior for the diagnostic accuracy of the reference test is investigated. In the real-data application, the proposed model was compared with the generally used logistic-regression model that ignores the imperfectness of the reference test. Conditional on the selected sample size and prior distributions, the simulation study results indicate satisfactory performance of the model-based estimates. In particular, the obtained average estimates for all parameters are close to the true values. For the real-data application, AUC estimates for the proposed model are substantially higher than those from the 'traditional' logistic-regression model.

show abstract

Predicting long-term risk for relationship dissolution using nonparametric conditional survival trees.

Kliem

Weusthoff

Hahlweg

et al. 2015

Journal of Family Psychology

View full text Add to dashboard Cite

Identifying risk factors for divorce or separation is an important step in the prevention of negative individual outcomes and societal costs associated with relationship dissolution. Programs that aim to prevent relationship distress and dissolution typically focus on changing processes that occur during couple conflict, although the predictive ability of conflict-specific variables has not been examined in the context of other factors related to relationship dissolution. The authors examine whether emotional responding and communication during couple conflict predict relationship dissolution after controlling for overall relationship quality and individual well-being. Using nonparametric conditional survival trees, the study at hand simultaneously examined the predictive abilities of physiological (systolic and diastolic blood pressure, heart rate, cortisol) and behavioral (fundamental frequency; f0) indices of emotional responding, as well as observationally coded positive and negative communication behavior, on long-term relationship stability after controlling for relationship satisfaction and symptoms of depression. One hundred thirty-six spouses were assessed after participating in a randomized clinical trial of a relationship distress prevention program as well as 11 years thereafter; 32.5% of the couples' relationships had dissolved by follow up. For men, the only significant predictor of relationship dissolution was cortisol change score (p = .012). For women, only f0 range was a significant predictor of relationship dissolution (p = .034). These findings highlight the importance of emotional responding during couple conflict for long-term relationship stability.

show abstract

A Smooth ROC Curve Estimator Based on Log-Concave Density Estimates

Cited by 7 publications

References 42 publications

Smooth time-dependent receiver operating characteristic curve estimators

Smooth time-dependent receiver operating characteristic curve estimators

Development of a diagnostic test based on multiple continuous biomarkers with an imperfect reference test

Predicting long-term risk for relationship dissolution using nonparametric conditional survival trees.

Contact Info

Product

Resources

About