Please use this identifier to cite or link to this item:
|Title:||Alternative Summary Indices: PLC and ASC for the Summary Receiver Operating Charcteristic (SROC) Curve|
|Keywords:||ROC curve;SROC curve;meta-analysis;projected length of the curve;area swept out by the curve;homogeneous|
|Abstract:||The receiver operating characteristic (𝘙𝘖𝘊) curve has been recommended to represent the performance of a diagnostic test. For decades many new methods and theory of 𝘙𝘖𝘊 curve have been developed. Among them the summary indices: the projected length of the 𝘙𝘖𝘊 curve (𝘗𝘓𝘊) and the area swept out by the 𝘙𝘖𝘊 curve (𝘈𝘚𝘊) firstly proposed by Lee provoked interests. More recently meta-analytic methods for diagnostic test accuracy have been developed and the summary receiver operating characteristic (𝘚𝘙𝘖𝘊) curve has been recommended to represent the performance of a diagnostic test based on the meta-analysis. The basic properties of the 𝘚𝘙𝘖𝘊 curve were discussed by Walter (2002). Based on his findings on the 𝘚𝘙𝘖𝘊 curve, my project focuses on studying the summary indices 𝘗𝘓𝘊 and 𝘈𝘚𝘊 in the context of 𝘚𝘙𝘖𝘊 curve. The mathematical expressions of these two indices and their variances are derived firstly in my project. Then the main characteristics of these two indices are discussed by resorting to software R codes in details. We find that 𝘗𝘓𝘊 index is symmetric with respect to parameters |α| and |𝘣| (α and 𝘣 are the coefficients of the regression model proposed by Moses 𝘦𝘵 𝘢𝘭. (1993)); affected mostly by the new proposed index M, 𝘈𝘚𝘊 index lacks symmetry with respect to 𝘣 for fixed α or lacks symmetry with respect to α for fixed 𝘣. The homogeneous values of 𝘗𝘓𝘊 provide a good approximation in heterogeneous studies. 𝘗𝘓𝘊 index gains its maximum √2 in the heterogeneous case. But as α -> ∞ attains its maximum 0.5 in the homogeneous case. Our example illustrates that the heterogeneous variances of both 𝘗𝘓𝘊 and 𝘈𝘚𝘊 are larger than the homogeneous estimates and the big values of var(𝘗𝘓𝘊)[L hat] and var(𝘈Ŝ𝘊) both indicate the worst situation for estimating these two indices in the diagnostic test.|
|Appears in Collections:||Digitized Open Access Dissertations and Theses|
Files in This Item:
|zhang_xuan_2004Dec_masters.pdf||2.45 MB||Adobe PDF||View/Open|
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.