You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
function [A,Aci] = auc(ego,data,alpha,flag,nboot,varargin);
665
-
% [A,Aci] = auc(data,alpha,flag,nboot,varargin);
666
-
%
667
-
% INPUTS
668
-
% data - Nx2 matrix [t , y], where
669
-
% t - a vector indicating class value (>0 positive class, <=0 negative)
670
-
% y - score value for each instance
671
-
%
672
-
% OPTIONAL
673
-
% alpha - level for confidence intervals (eg., enter 0.05 if you want 95% CIs)
674
-
% flag - 'hanley' yields Hanley-McNeil (1982) asymptotic CI
675
-
% 'maxvar' yields maximum variance CI
676
-
% 'mann-whitney'
677
-
% 'logit'
678
-
% 'boot' yields bootstrapped CI (DEFAULT)
679
-
% nboot - if 'boot' is set, specifies # of resamples, default=1000
680
-
% varargin - additional arguments to pass to BOOTCI, only valid for 'boot'
681
-
% this assumes you have the STATs toolbox, otherwise it's
682
-
% ignored and a crude percentile bootstrap is estimated.
683
-
%
684
-
% OUTPUTS
685
-
% A - area under ROC
686
-
% Aci - confidence intervals
687
-
%
688
-
% EXAMPLES
689
-
% % Classic binormal ROC. 100 samples from each class, with a unit mean separation
690
-
% % between the classes.
691
-
% >> mu = 1;
692
-
% >> y = [randn(100,1)+mu ; randn(100,1)];
693
-
% >> t = [ones(100,1) ; zeros(100,1)];
694
-
% >> [A,Aci] = auc([t,y])
695
-
% >> trueA = normcdf(mu/sqrt(1+1^2))
696
-
%
697
-
% REFERENCE
698
-
% Gengsheng Qin & Lejla Hotilovac. Comparison of non-parametric
699
-
% confidence intervals for the area under the ROC curve of a continuous
700
-
% scale diagnostic test.
701
-
% Stat Methods Med Res 17:207, 2008
702
-
671
+
%> @param data - [class , score] matrix
672
+
%> @param alpha - level for confidence intervals (eg., enter 0.05 if you want 95% CIs)
673
+
%> @param flag - 'hanley' yields Hanley-McNeil (1982) asymptotic CI; 'maxvar' yields maximum variance CI;'mann-whitney';'logit';'boot' yields bootstrapped CI (DEFAULT)
674
+
%> @param nboot - if 'boot' is set, specifies # of resamples, default=1000
675
+
%> @param varargin - additional arguments to pass to BOOTCI, only valid for 'boot' this assumes you have the STATs toolbox, otherwise it's ignored and a crude percentile bootstrap is estimated.
0 commit comments