June 18th 2015
Faculty of Economic and Business Sciences | Seminar 105
Nuevas Medidas para valorar Biomarcadores
2015/06/18 – 11:30 h | M.C. Pardo, Department of Statistic & Operation Research, Complutense University of Madrid (UCM)
María del Carmen Pardo was born in Segovia, Spain, in 1969. In 1992 she graduated in Mathematics in the field of Operational Research at the Complutense University of Madrid. She continued her postgraduate studies there where she received the Ph. D. degree in Statistics in 1996. From 1992-1994, she was Assistant Professor and then Associate Professor at the Complutense University of Madrid. During years, her research interest was the use of measures of information theory in statistical inference. Nowadays, her research line is Biostatistics, in particular survival analysis, longitudinal data and assessment of diagnostic tests. She is co-chair of the specialized teamStatistical Analysis of Event Times of ERCIM WG on Computational and Methodological Statistics. She has been a visiting scientist for several times at the Institute of Information Theory and Automation, Czech Academy of Sciences, Prague, McMaster University, Canada and Miguel Hernández University of Elche, Spain.
She has been awarded the “Best Graduate Prize” for obtaining the best marks in the field of Operational Research as well as “Best Ph. D. Prize” to the best Ph. D. Dissertation in Statistics and Operational Research. Also, she was elected to participate in the 10th European Young Statisticians Meeting and she has been awarded the “Ramiro Melendreras Prize” from the Spanish Society of Statistical and Operational Research.
Abstract
En medicina la curva ROC permite valorar la capacidad de un biomarcador para discriminar entre individuos sanos y enfermos (Pepe (2003)). La forma usual de comparar dos biomarcadores es utilizar el estadístico de DeLong et al. (1988) basado en el área bajo la curva (AUC) receiver operating characteristic (ROC) de ambos biomarcadores. Sin embargo, en ocasiones se puede dar el caso de que diferentes ROCs tengan un valor AUC parecido y que este procedimiento falle en distinguir las cuvas. Han sido varios los estadísticos propuestos como alternativa a este procedimiento como por ejemplo el de Nakas et al. (2003). Sin embargo, la mayoría no pueden escribirse en función de la ROC por lo que pierden la interpretación geométrica que tiene el estadístico de DeLong y no son tan populares.
En esta charla se presenta una medida basada en la ROC empírica pero que distingue ROCs con AUCs similares. Se presenta un amplio estudio de simulación para analizar el nuevo procedimiento de valoración de biomarcadores en comparación con los estadísticos mencionados anteriormente. Finalmente, se analiza un conjunto de datos reales.
References:
DeLong, E.R., DeLong, D.M. & Clarke-Pearson, D.L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach, Biometrics 44: 837-845.
Nakas, C., Yiannoutsos, C.T., Bosch & R.J. Moyssiadis, C. (2003). Assessment of diagnostic markers by goodness-of-fit tests. Statistics in Medicine. 22, 2503-2513.
Pepe, M.S. (2003). The Statistical Evaluation of Medical Tests for Classification and Prediction, Oxford Statistical Science Series, Oxford University Press.
Gallery
