Find observations where three or more numeric variables exceed values

[eh51]

My solution was to go with proc ginside. 

However, I broke the 4 dimensional problem into a set of 10 2 dimensional problems.

I separated my limits data set into 10 data sets based on the 'temp' and 'cat' variables. 

I also separated my operational data into 10 data sets also by 'temp' and 'cat' variables. 

I then used proc ginside to determine which observations were inside and which were outside the limit polygon.  

I then concatenated the 10 sets back to one.  

Thanks for the advice. 

The code is basic, but is here for benefit of readers. Feedback is welcome also. 

/*   Create the boundary tables    */

data lim_35090 (keep=x y limid);
set sasuser.limits2330 (rename=(speed=x weight=y));
where(cat= 0.9  & temperature=35);
run;

data lim_35095 (keep=x y limid);
set sasuser.limits2330 (rename=(speed=x weight=y));
where(cat= 0.95 & temperature=35);

...8 others similar

/*  Create tables of the operating data      */ 
DATA ops_30090 ops_30095 ops_30100 ops_30105 ops_30110 ops_35090 ops_35095 ops_35100 ops_35105 ops_35110;
    SET op_base1 (rename=(speed=x weight=y));
	if temperature <= 30 then do;
			if cat <= 0.90 THEN OUTPUT ops_30090;
			else if (0.90 < cat <= 0.95) THEN OUTPUT ops_30095;
			else if (0.95 < cat <= 1) THEN OUTPUT ops_30100;
			else if (1 < cat <= 1.05) THEN OUTPUT ops_30105;
			else OUTPUT ops_30110;
		end;	
	
	else do;
			if cat <= 0.90 THEN OUTPUT ops_35090;
			else if (0.90 < cat <= 0.95) THEN OUTPUT ops_35095;
			else if (0.95 < cat <= 1) THEN OUTPUT ops_35100;
			else if (1 < cat <= 1.05) THEN OUTPUT ops_35105;
			else OUTPUT ops_35110;
		end;
RUN;

/* macro to evaluate if observation is inside limits*/
OPTIONS MPRINT;
%macro insout(suffix=);
	
	proc ginside map=lim_&suffix data=ops_&suffix
	     out=results_&suffix
	     includeborder
	  ;
	  id limid;
	run;
			
%mend insout;

/*  Run macros for ginside  */
%insout(suffix=30090);
%insout(suffix=30095);
%insout(suffix=30100);
%insout(suffix=30105);
%insout(suffix=30110);
%insout(suffix=35090);
%insout(suffix=35095);
%insout(suffix=35100);
%insout(suffix=35105);
%insout(suffix=35110);

[Ksharp]

Or you could try Principle Component Analysis to map these three variable into X and Y axis ,
and use PROC GINSIDE or Euclidean/Mahalanobis distance ?

https://blogs.sas.com/content/iml/2014/11/07/distribution-of-blood-types.html