Two-dimension PDF made by PROC SGPLOT to estimate the probability mass

[TomHsiung]

Hi, guys

I expand my research of PDF graph to two variables. Here is my code.

data Have; set Sample2(keep=weight inf_weight); run;
   proc kde data=Have;
     bivar weight inf_weight / bwm=2 ngrid=250 plots=contour;
     score data=Have out=Out(rename=(value1=weight value2=inf_weight));
   run;
   proc sort data=Out;  by density;  run;
   proc sgplot data=Out;
      styleattrs wallcolor=white;
      scatter x=weight y=inf_weight / colorresponse=density colormodel=ThreeColorRamp
                         markerattrs=(symbol=CircleFilled) transparency=0.5;
      xaxis grid values=(30 to 150); 
      yaxis grid values=(500 to 6000);
   run;

This will produce a scatter plot and the value of PDFs is represented by the color.

If I intend to know the area that corresponds to a probability mass of 95%, how do I get it and mark the area on the graph? I know there is another type of graph called surface plot which uses the volume to represent the probability mass but that's not my consideration. Thanks.

[Ksharp · Accepted solution]

Check @Rick_SAS  blogs:

https://blogs.sas.com/content/iml/2014/07/21/add-prediction-ellipse.html

https://blogs.sas.com/content/iml/2014/07/23/prediction-ellipses-from-covariance.html

data Have; set sashelp.heart(keep=weight height); run;
   proc kde data=Have ;
     bivar weight height /out=out(rename=(value1=weight value2=height)) bwm=2 ngrid=250  plots=contour;
   run;
   proc sort data=Out;  by density;  run;

   data out2;
    set out have(keep=weight height rename=(weight=weight2 height=height2));
run;


   proc sgplot data=Out2;
      styleattrs wallcolor=white;
      scatter x=weight y=height / colorresponse=density colormodel=ThreeColorRamp
                         markerattrs=(symbol=CircleFilled) transparency=0.5;
	   ellipse x=weight2 y=height2;           /* default is ALPHA=0.05 */

   run;

[TomHsiung]

Thanks, dear buddy @Ksharp