Writing custom probbnrm function

[SergeSU]

Hello!

I am trying to write a SAS (macro) function that returns a value, in an effort to create a PROBBNRM function analog that allows customized lower/upper bounds and accepts non-standardized bivariate normal distribution means/variances.

I have written a macro function (I think) that computes the desired probability from user input,

%JointProbability(mu1, mu2, sigma1, sigma2, corr, critx1, critx2, crity1, crity2);quit;

The code for the macro is below, but all it does is assign the resulting probability to a global variable MyOutput, rather than return a value. I intend to use this function in proc nlp to fit a bivariate model (Platform: Windows 2K, SAS 9 used).

Would anyone have any suggestions on how to proceed?

Thank you!!
Serge


%macro JointProbability(xMean,yMean,xSD,ySD,rho,xLower,xUpper,yLower,yUpper);
data;
m1=&xMean;m2=&yMean;
s1=&xSD;s2=&ySD;
r=ρ
cx1=&xLower;cx2=&xUpper;
cy1=&yLower;cy2=&yUpper;
zx1=(cx1-m1)/s1;zx2=(cx2-m1)/s1;zx3=(cx1-m1)/s1;zx4=(cx2-m1)/s1;
zy1=(cy1-m2)/s2;zy2=(cy2-m2)/s2;zy3=(cy2-m2)/s2;zy4=(cy1-m2)/s2;
p_out=probbnrm(zx2,zy2,r)-probbnrm(zx3,zy3,r)-probbnrm(zx4,zy4,r)+probbnrm(zx1,zy1,r);
%global MyOutput;
call symput('MyOutput',p_out);
run;
%mend;

 

[ChrisW75]

First off, that macro will end your datastep if you run it within a datastep as it has a data statement and run statement in it. You should remove those.

You should be able to use this as an example of what you can do...

%macro test();
  %let mvar = tester;
  &mvar;
%mend test;

data _null_;
  new_var = "%test";

  put new_var=;
run;

new_var=tester

As you can see, putting the macro variable on line there returns the value. You can do the same ting without the quotes for a numeric value.

%macro test();
  %let mvar = 5;
  &mvar;
%mend test;

data _null_;
  new_var = %test;

  put new_var=;
run;

new_var=5;

Enjoy.

Chris
Business Analyst, Code Monkey, Data Wrangler.
SAS Guru.