Regression Models for Convex ROC Curves

Lloyd, Chris J.

doi:10.1111/j.0006-341x.2000.00862.x

Cited by 17 publications

(14 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Some trials to improve this estimator are presented in detail, for example, in Lloyd (2002) or Horová et al (2012).…”

Section: Estimation Of the Roc Curvementioning

confidence: 99%

Empirical and Kernel Estimation of the ROC Curve

Baszczyńska

2015

Folia Oeconomica

View full text Add to dashboard Cite

The paper presents chosen methods for estimating the ROC (Receiver Operating Characteristic) curve, including parametric and nonparametric procedures. Nonparametric approach may involve the use of empirical method or kernel method of the ROC curve estimation. In the analysis, an attempt of comparison of empirical and kernel ROC estimators is done, considering the impact of sample size, choice of smoothing parameter and kernel function in kernel estimation on the results of the estimation. Based on the results of simulation studies, some suggestions, useful in the procedures of nonparametric ROC curve are determined.

show abstract

“…Some trials to improve this estimator are presented in detail, for example, in Lloyd (2002) or Horová et al (2012).…”

Section: Estimation Of the Roc Curvementioning

confidence: 99%

Empirical and Kernel Estimation of the ROC Curve

Baszczyńska

2015

Folia Oeconomica

View full text Add to dashboard Cite

show abstract

“…Average management costs are minimized at the point on the ROC curve (22) with slope (FPP) f ′ (49). Methodology for identifying this point has been described fully in Hughes and Madden (14) and Madden (23), and is not presented in detail here.…”

Section: Predictor Variable (Type) Descriptionmentioning

confidence: 99%

“…Methodology for identifying this point has been described fully in Hughes and Madden (14) and Madden (23), and is not presented in detail here. Briefly, the empirical ROC curves were expressed in a functional (parametric) form using the equation of Lloyd (22). Shape parameters of the parametric ROC curves, defined as Δ and µ, were estimated by nonlinear regression in SAS following equation 4 of Turechek and Wilcox (46).…”

Section: Predictor Variable (Type) Descriptionmentioning

confidence: 99%

Site-Specific Risk Factors for Ray Blight in Tasmanian Pyrethrum Fields

et al. 2009

View full text Add to dashboard Cite

Ray blight of pyrethrum (Tanacetum cinerariifolium), caused by Phoma ligulicola var. inoxydablis, can cause defoliation and reductions of crop growth and pyrethrin yield. Logistic regression was used to model relationships among edaphic factors and interpolated weather variables associated with severe disease outbreaks (i.e., defoliation severity ≥40%). A model for September defoliation severity included a variable for the product of number of days with rain of at least 0.1 mm and a moving average of maximum temperatures in the last 14 days, which correctly classified (accuracy) the disease severity class for 64.8% of data sets. The percentage of data sets where disease severity was correctly classified as at least 40% defoliation severity (sensitivity) or below 40% defoliation severity (specificity) were 55.8 and 71%, respectively. A model for October defoliation severity included the number of days with at least 1 mm of rain in the past 14 days, stem height in September, and the product of the number of days with at least 10 mm of rain in the last 30 days and September defoliation severity. Accuracy, sensitivity, and specificity were 72.6, 73.6, and 71.4%, respectively. Youden's index identified predictive thresholds of 0.25 and 0.57 for the September and October models, respectively. When economic considerations of the costs of false positive and false negative decisions and disease prevalence were integrated into receiver operating characteristic (ROC) curves for the October model, the optimal predictive threshold to minimize average management costs was 0 for values of disease prevalence greater than 0.2 due to the high cost of false negative predictions. ROC curve analysis indicated that management of the disease should be routine when disease prevalence is greater than 0.2. The models developed in this research are the first steps toward identifying and weighting site and weather disease risk variables to develop a decision-support aid for the management of ray blight of pyrethrum. RightsWorks produced by employees of the U.S. Government as part of their official duties are not copyrighted within the U.S. The content of this document is not copyrighted. Pyrethrum (Tanacetum cinerariifolium) is a perennial plant in the family Asteraceae that is cultivated for the production of six closely related esters called pyrethrins, which possess insecticidal properties (7,10,48). Pyrethrum is an integral component of the cropping system rotation used in the northwest coast of Tasmania, contributing approximately 30% of the world's supply of pyrethrins. The crop is planted by seed from July to September, and harvests are conducted annually for the duration of the crop (typically 4 to 5 years) (34). AuthorsIn the absence of fungicide applications in spring, ray blight, a fungal disease caused by Phoma ligulicola var. inoxydablis, can reduce plant growth and pyrethrin yield in Tasmania (31,34,36,40). Ray blight was first observed causing minor losses on pyrethrum flowers in Tasmania in 1995 (41). Since 1999, c...

show abstract

“…An early reference is Swets and Pickett (1982), more recent approaches to estimation are proposed in (Lloyd, 2000) and (Venkatraman, 2000). …”

Section: Receiver Operating Characteristicmentioning

confidence: 99%

Identification of interaction patterns and classification with applications to microarray data

Boulesteix¹,

Tutz²

2006

Computational Statistics & Data Analysis

View full text Add to dashboard Cite

Emerging patterns represent a class of interaction structures which has been recently proposed as a tool in data mining. In this paper, a new and more general definition refering to underlying probabilities is proposed. The defined interaction patterns carry information about the relevance of combinations of variables for distinguishing between classes. Since they are formally quite similar to the leaves of a classification tree, we propose a fast and simple method which is based on the CART algorithm to find the corresponding empirical patterns in data sets. In simulations, it can be shown that the method is quite effective in identifying patterns. In addition, the detected patterns can be used to define new variables for classification. Thus, we propose a simple scheme to use the patterns to improve the performance of 1 classification procedures. The method may also be seen as a scheme to improve the performance of CARTs concerning the identification of interaction patterns as well as the accuracy of prediction.2

show abstract

Regression Models for Convex ROC Curves

Cited by 17 publications

References 28 publications

Empirical and Kernel Estimation of the ROC Curve

Empirical and Kernel Estimation of the ROC Curve

Site-Specific Risk Factors for Ray Blight in Tasmanian Pyrethrum Fields

Identification of interaction patterns and classification with applications to microarray data

Contact Info

Product

Resources

About