可采用medcalc计算

symbol1  v=      i= spline       .........

不知道你开始数据格式,就从开始的logistic生成sensitivity和1-specificity两个变量开始, 不知是不是你想要的

proc logistic data = psa;
       model capsule (event = "1") = gleason psa / outRoc = MyRocTable;/*Asks to put calculations for an ROC table into a new dataset, which we’re calling “MyRocTable” here*/
  title ' ';
run;
/* make an ROC plot*/
proc gplot data = MyRocTable;
        plot _sensit_  * _1mspec_ ;
run;
quit;

/* use a line instead of dots on the graph*/
proc gplot data = MyRocTable;
        symbol1 i=join v = none c=black line=1;
        plot _sensit_  * _1mspec_ ;
run;
quit;
/* use a Smo0th line instead of line on the graph,The greater the nn value, the smoother the fitted curve. By default, the value of nn is 0*/
proc gplot data = MyRocTable;
        symbol1 i=sm40 v = none c=black line=1;/*i=sm<nn><P><S>*/
        plot _sensit_  * _1mspec_ ;
run;
quit;

4# realgod5417
请问高手，利用symbol1 i=sm40 平滑后的曲线，能否从中得到平滑后的数据集？谢谢！

Smooth ROC curve is often from an approximate of some distributions, such as bi-normal. Thus, you can fit in some procedures, such as no-linear model, get the data set and then plot it. JingJu