Nonparametric Cutoff Point Estimation for Diagnostic Decisions with Weighted Errors

Estimación no paramétrica del punto de corte asociado a una decisión diagnóstica con errores ponderados

PABLO MARTÍNEZ-CAMBLOR1

1Oficina de Investigación Biosanitaria, CAIBER, Oviedo, Spain. Universidad de Oviedo, Departamento de Estadística e I.O. y D.M., Oviedo, Spain. Biostatistics and Associate professor. Email: pablomc@ficyt.es


Abstract

The study of diagnostic tests is a hot topic which has direct applications in biomedical sciences. Despite of the relevance, in a diagnostic process, of the threshold (or cutoff point) employed on the decision taken by the physician, the study and comparison of the accuracy among different diagnostic criterions has been the main field of study. In this paper, the authors are interested in the study of the involved cutoff point estimation in diagnostic tests with weighted errors. With this goal, a nonparametric smoothed utility function estimator is considered. The bootstrap and the asymptotic distributions for the related M-estimator are derived. Finally, the obtained results are applied to study the Procalcitonin level which determines whether a child within the Pediatric Intensive Care Unit (UCIP) has a virical sepsis.

Key words: Kernel density estimator, Sensitivity, Specificity, Threshold, Utility function.


Resumen

El estudio de tests diagnósticos es un tema candente con aplicaciones directas en las ciencias biomédicas. Aunque en la práctica, a la hora de tomar una decisión, los clínicos deben fijar un valor umbral (o punto de corte) a pesar de la relevancia que este valor tiene, el estudio y la comparación de la calidad entre diferentes criterios diagnósticos ha sido el principal campo de estudio. En este trabajo, los autores están interesados en el estudio de la estimación del punto de corte involucrado en un test diagnóstico con errores ponderados. Con este objetivo, se considera un estimador suavizado para una función de utilidad. Se estudian las distribuciones bootstrap y asintóticas del M-estimador resultante. Finalmente, los resultados obtenidos son aplicados al estudio de los niveles de Procalcitonina que determinan si un niño ingresado en la Unidad de Cuidados Intensivos Pediátricos (UCIP) tiene infección vírica.

Palabras clave: especificadas, estimador núcleo para la densidad, función de utilidad, sensibilidad, umbral.


Texto completo disponible en PDF


References

1. Airola, A., Pahikkala, T., Waegeman, W., De Baets, B. & Salakoski, T. (2011), `An experimental comparison of cross-validation techniques for estimating the area under the ROC curve´, Computational Statistics & Data Analysis 55(4), 1828-1844.

2. Cao, R. (1990), Aplicaciones y nuevos resultados del Método Bootstrap en la estimación no paramétrica de curvas, PhD thesis, University of Santiago de Compostela, Santiago de Compostela, Spain.

3. Chin-Ying, L., Tian, L. & Schisterman, E. F. (2011), `Exact confidence interval estimation for the Youden index and its corresponding optimal cut-point´, Computational Statistics & Data Analysis. In Press, Corrected Proof. DOI: 10.1016/j.csda.2010.11.023. *http://www.sciencedirect.com/science/article/B6V8V-51N223Y-1/2/9ac9b12dcddcf280e599a152375ca56c

4. Efron, B. & Tibshirani, R. J. (1993), An Introduction to the Bootstrap, Chapman and Hall, London, United Kingdom.

5. Fluss, R., Faraggi, D. & Reiser, B. (2005), `Estimation of the Youden index and its associated cutoff point´, Biometrical Journal 47(4), 458-472.

6. Green, D. M. & Swets, J. A. (1966), Signal Detection Theory and Psychophysics, Wiley, New York, United States.

7. Hall, P., DiCiccio, J. T. & Romano, J. P. (1989), `On smoothing and the bootstrap´, Annals of Statistics 17, 692-702.

8. Hand, D. J. (2009), `Measuring classifier performance: A coherent alternative to the area under the ROC curve´, Machine Learning - ML 77(1), 103-123.

9. Huang, J. & Ling, C. X. (2005), `Using AUC and accuracy in evaluating learning algorithms´, IEEE Transactions on Knowledge and Data Engineering 17(3), 299-310.

10. Krzanowski, W. J. & Hand, D. J. (2009), ROC Curves for Continuous Data, Chapman and Hal, New York, United States.

11. Lee, W. C. & Hsiao, C. K. (1998), `Alternative summary indices for the receiver operating characteristic curve´, Epidemiology 7, 605-611.

12. Liu, R. & Yang, L. (2008), `Kernel estimation of multivariate cumulative distribution function´, Journal of Nonparametric Statistics 20(8), 661-667.

13. López-de-Ulibarri, I., Cao, R., Cadarso-Suárez, C. & Lado, M. J. (2008), `Nonparametric estimation of conditional ROC curves: application to discrimination tasks in computarized detection of early breast cancer´, Computational Statistics & Data Analysis 52(5), 2623-2631.

14. Martínez-Camblor, P. & De Uña-Álvarez, J. (2009), Studying the bandwidth in k-sample smooth tests, Technical Report 09, Universidad de Vigo, Vigo, Spain.

15. Martínez-Camblor, P. & Yáñez-Juan, A. (2009), `Testing the equality of diagnostic effectiveness of one measure with respect to k different features´, Journal of Applied Statistics 36(4), 359-367.

16. Mugdadi, A. R. & Ghebregiorgis, G. S. (2005), `The kernel distribution estimator of functions of random variables´, Journal of Nonparametric Statistics 17(7), 807-818.

17. Nadaraya, E. A. (1962), `Some new estimates for distribution functions´, Theory Probability Application 9, 497-500.

18. Rey, C., Los Arcos, M., Concha, A., Medina, A., Prieto, S., Martínez-Camblor, P. & Prieto, B. (2007), `Procalcitonin and C-reactive protein as markers of systemic inflammatory response syndrome severity in critically ill children´, Intensive Care Medicine 33(3), 477-484.

19. Rodríguez-Álvarez, M. X., Tahoces, P. G., Cadarso-Suárez, C. & Lado, M. J. (2011), `Comparative study of ROC regression techniques-applications for the computer-aided diagnostic system in breast cancer detection´, Computational Statistics & Data Analysis 55(1), 888-902.

20. Rosenblatt, M. (1956), `Remarks on some nonparametric estimates of a density function´, Annals of Mathematical Statistics 17, 832-837.

21. Silverman, B. W. (1978), `Weak and strong uniform consistency of the density estimation and its derivatives´, Annals of Statistics 6, 1177-1184.

22. Silverman, B. W. (1986), Density Estimation for Statistics and Data Analysis, Chapman and Hall, New York, United States.

23. Youden, W. J. (1950), `Index for rating diagnostic test´, Cancer 3, 32-35.


[Recibido en junio de 2010. Aceptado en diciembre de 2010]

Este artículo se puede citar en LaTeX utilizando la siguiente referencia bibliográfica de BibTeX:

@ARTICLE{RCEv34n1a07,
    AUTHOR  = {Martínez-Camblor, Pablo},
    TITLE   = {{Nonparametric Cutoff Point Estimation for Diagnostic Decisions with Weighted Errors}},
    JOURNAL = {Revista Colombiana de Estadística},
    YEAR    = {2011},
    volume  = {34},
    number  = {1},
    pages   = {133-146}
}